Program / Abstracts www.sbmc2016.de 1 Program / Abstracts 2 CONTENT CONTENT 3 WELCOME5 ORGANIZER 8 GENERAL INFORMATION 9 SOCIAL PROGAM 11 PROGRAM OVERVIEW 13 PROGRAM 14 POSTER OVERVIEW 18 MTZ AWARD 26 INVITED TALKS 29 ORAL PRESENTATIONS 35 POSTER PRESENTATIONS 47 INDEX 123 3 Program / Abstracts 4 WELCOME Message of greeting for the 6th International Conference on Systems Biology of Mammalian Cells (SBMC 2016) Major diseases such as cancer and dementia are caused by specific dysfunctions in complex, dynamic life processes of human cells and organs. Many genetic factors as well as environmental influences and individual lifestyles have a decisive effect on the development and progress of such diseases as well as on the patient’s prospects. Systems biology and systems medicine, which derives from it, make it possible to understand the complex connections and thus to improve patient care in the long term. The Federal Ministry of Education and Research has been supporting systems medicine research through the research and funding scheme “e:Med – Paving the Way for Systems Medicine” since 2012. The latter is part of the Federal Government’s Health Research Framework Programme. Along with medical genome research, systems biology is the main contributor to progress in systems medicine. German research has a pioneering role in this field in Europe and we must continue to strengthen it. My Ministry will continue to expand national expertise, support international collaborations and promote young research talent. The exchange of experience between researchers often contributes substantially to advances in research. The Conference on Systems Biology of Mammalian Cells 2016 provides an excellent opportunity for scientists to share their insights and make new contacts. It also provides the opportunity to learn about developments in systems biology and systems medicine research. I am particularly glad that the conference will be recognizing the work of young scientists: The doctoral theses which will be honoured with the “MTZ Award for Medical Systems Biology” of the MTZ Foundation clearly show the potential of young scientists in the field of medically oriented systems biology. I wish all the participants interesting discussions and many new ideas for their scientific work. Prof. Dr. Johanna Wanka Federal Minister of Education and Research 5 Program / Abstracts WELCOME As the CEO of the Helmholtz Zentrum München I am pleased to welcome you to the 6th Conference on Systems Biology of Mammalian Cells (SBMC 2016) in Munich. The Helmholtz Zentrum München has a unique profile by focusing on major common diseases such as diabetes mellitus, chronic lung diseases and in the near future additionally on allergies. Our goal is to contribute to personalized approaches for the prevention and therapy of those common diseases thus causing an impact on the patient’s health. To achieve this, the scientists investigate the interaction of genetics, environmental factors and lifestyle with new technologies in different areas such as epidemiology, epigenetic and stem cells. The most exciting new project though is the Pioneer Campus. The Pioneer Campus will offer brilliant young scientists in interdisciplinary teams from around the world a research facility in a highly creative, excellent academic and outstanding technological environment. The central vision of the Pioneer Campus is to enable decisive breakthroughs in the development of concepts for key technological innovations. System biology - I am sure - will also play an important role in this concept of the Pioneer Campus as it is not only an enabler but also a driver in itself for development of new hypotheses. Systems biology has therefore a long tradition at the Helmholtz Zentrum München. In 2008, the interdisciplinary project ‘Control of Regulatory Networks’ was begun within the ‘Alliance for Systems Biology’ funded by the Helmholtz Association. Today, there are many experimental and theoretical approaches in systems biology at our Center as well as in the larger context of the research field “Health” of the Helmholtz Alliance. That is why we are glad to organize this year’s SBMC here in Munich. At SBMC 2016, systems biology methods and their applications in various fields of basic research will be presented along with translation. Thus the conference contributes as a platform for the presentation and the exchange of the latest systems biology research into important social objectives: the elucidation of disease mechanisms and the development of personalized medicine. Considering all these facts, I am sure the research we are doing will have an important and positive impact on human health. I hope that you will enjoy the conference and wish you successful days. rof. Dr. Günther Wess P CEO Helmholtz Zentrum München 6 WELCOME Dear colleagues and friends, With great pleasure we welcome you today to the 6th Conference on Systems Biology of Mammalian Cells here in Munich. This conference is kindly supported by the BMBF, the German Federal Ministry of Education and Research and is a continuation of a conference series started in 2006, organized early on by the competence network HepatoSys, which 4 years later was successfully extended by the Virtual Liver Network promoting excellent interdisciplinary research. In the beginning of 2016, the BMBF granted the follow-up competence network LiSym, enabling further progress of systems biology towards applied systems medicine in the context of liver research. Systems biology has arrived, and many experimental groups have started to incorporate corresponding data analyses and modeling approaches into their daily work flows. Similarly, theoretical groups have either taken up biological experiments, or have started intense collaboration with experimental partners with daily interactions. Next steps are towards integration of ever increasing, ever more complex data as well as available information from existing resources into models, from basic research to translational systems medicine. SBMC 2016 will present such novel approaches progress, gathering outstanding theoretical and practical scientists from all over the world with the intention to promote exchange of ideas and knowledge for further acceleration of advances of systems biology. The program was designed to cover a broad range of theoretical, experimental and computational approaches and technology trends including highlights in the field of novel technologies like single-cell profiling and image-based systems biology as well as scientific progress in modeling approaches like signaling and metabolism and genetic/epigenetic mechanisms. Moreover, progress in systems medicine and pharmacology complete the compilation of cutting-edge research and excellent progress in theoretical and applied systems biology. The outstanding list of invited speakers include top international researchers like our plenary speaker Chris Sander from Harvard Medical School and many more. Moreover, we are glad to host the MTZ-Award ceremony 2016 for excellent PhD theses in systems biology, with three young scientists presenting their innovative research to the SBMC audience. The SBMC program includes two poster sessions, which encourage detailed one-on-one interactions, as well as a welcome reception on the first day and a conference dinner/ party at the BMW World in Munich. We are sure you will enjoy this year´s SBMC in Munich and we are looking forward to many fruitful discussions, productive interactions and exciting after-conference activities here at Munich. Thanks to all those who have contributed to enabling and organizing SBMC 2016! Fabian Theis Ursula Klingmüller 7 Program / Abstracts ORGANIZER Conference Chairs Fabian Theis Institute of Computational Biology Helmholtz Zentrum München Deutsches Forschungszentrum für Gesundheit und Umwelt (GmbH) Ingolstädter Landstr. 1 85764 Neuherberg Ursula Klingmüller Systems Biology of Signal Transduction German Cancer Research Center (DKFZ) Im Neuenheimer Feld 280 69120 Heidelberg In Cooperation with: Virtual Liver Network / LiSyM The 6th Conference on Systems Biology of Mammalian Cells is under the auspices of Prof. Dr. Johanna Wanka, Federal Minister of Education and Research Scientific Committee Dirk Drasdo (INRIA French National Institute for Research in Computer Science and Control, Le Chesnay Cedex) Jan Hengstler (Leibniz-Institut für Arbeitsforschung an der TU Dortmund, Dortmund) Thomas Höfer (DeutschesKrebsforschungszentrum (DKFZ), Heidelberg) Hermann-Georg Holzhütter (Charité - Universitätsmedizin Berlin, Berlin) Ursula Klingmüller (Deutsches Krebsforschungszentrum (DKFZ), Heidelberg) Frank Lammert (Universitätsklinikum des Saarlandes, Homburg/Saar) Madlen Matz-Soja (Universität Leipzig, Medizinische Fakultät, Leipzig) Werner Mewes (Helmholtz Zentrum München, Deutsches Forschungszentrum für Gesundheit und Umwelt (GmbH), Neuherberg) Nassir Navab (Technische Universität München, München) Ivo Sbalzarini (Max Planck Institute of Molecular Cell Biology and Genetics, Dresden) Heribert Schunkert (Deutschen Herzzentrum München, Lübeck) Fabian Theis (Helmholtz Zentrum München, Deutsches Forschungszentrum für Gesundheit und Umwelt (GmbH), Neuherberg) Jens Timmer (Freiburg Institute for Advanced Studies, School of Life Sciences – LifeNet, Freiburg im Breisgau) Nachiket Vartak (Leibniz-Institut für Arbeitsforschung an der TU Dortmund, Dortmund) Conference Office 8 event lab. GmbH Dufourstraße 15 04107 Leipzig Phone: Phone on site : Mail: 0341 / 240596-72 +49 177 2673158 sbmc2016@eventlab.org Plenary Speaker Chris Sander (cBio Center at Dana-Farber Cancer Institute and Harvard Medical School, Boston) Invited Speakers Alexander Anderson (H. Lee Moffitt Cancer Center & Research Institute, Tampa) Ina Bergheim (Friedrich Schiller University Jena, Jena) Johan Björkegren (Mount Sinai Hospital, New York) Bernd Bodenmiller (University of Zürich, Zürich) Javier Diego (EMBL-CRG, Barcelona) Alexander Hoffmann (UCLA, Los Angeles) Martin Howard (John Innes Centre, Norwich) Andrzej M. Kierzek (Simcyp, Sheffield) Markus Löffler (University of Leipzig, Leipzig) John Marioni (EBI, Hinxton) Robert Murphy (Carnegie Mellon University, Pittsburgh) Tudor Oprea (University of New Mexico, Albuquerque) David Rand (University of Warwick, Coventry) Ines Thiele (Esch-sur-Alzette, Luxembourg) GENERAL INFORMATION Conference Venue Klinikum rechts der Isar Technische Universität München Hörsaal A Ismaninger Straße 22 / Entrance Einsteinstraße 81675 München Germany You’ll find a map at the back. All lectures take place in the Hörsaal A. The exhibition and poster area and the registration desk can be found in the Foyer of Hörsaal A as well. Transport From Munich main station (about 10 min): From the main station take U4 (direction `Arabellapark’) or U5 (direction `Neuperlach Süd’) to`Max-Weber-Platz’ From the airport (about 40 min): Take S train S8 (direction `Herrsching’) to `Ostbahnhof’. Change to the underground U4 (direction `Laimer Platz’) or U5 (direction `Westendstrasse’) to `Max-Weber-Platz’ 9 Program / Abstracts Wifi Access The “eduroam” network is available at the university. Login details are available at the reception desk. Registration Desk The Registration Desk is located in the Foyer area. Opening hours of registration desk: Wednesday, April 06, 2016 12:00 noon – 08:00 pm Thursday, April 07, 2016 08:00 am – 07:00 pm Friday, April 08, 2016 08:00 am – 01:00 pm Technical Facilities / Speakers Preview The lecture room is equipped with laptop/data projectors and with blackboards. Speakers are requested to bring the file of their talks (preferably on a flash drive in PDF format) to the “speakers preview” at the registration desk, two hours before the beginning of their session, or to the technical support personnel in the lecture room not later than 30 minutes before the session starts. The “speakers preview” will be open during the opening hours of the registration desk. Poster Exhibition The poster exhibition of the 6th Conference on Systems Biology of Mammalian Cells will be located at the foyer area (level 0). The poster exhibition will be opened all-day on the conference days. The Poster Party, during which each presenter must be available for discussion, will take place on Wednesday, April 6th, at 6:30 pm (for even poster numbers) and Thursday, April 7th, at 6:00 pm (for odd poster numbers). Materials for putting up the posters will be provided. Posters must be removed from the board until the end of the conference. The meeting staff will remove posters not taken down by Friday, April 8th, and the meeting organizers cannot take any further responsibility for the material. We can proudly announce that the committee will award a prize to at least one outstanding poster. The poster prize ceremony will be part of the closing statement on Friday, April 8th. Conference Language The official language of the conference is English. 10 SOCIAL PROGAM Welcome Reception Do not miss the chance to mingle with colleagues and friends on the evening of April 6th, 2016 at 6:30 pm at the exhibition area in the Foyer. Tickets are included in the registration fee, but registration is required. Conference Dinner and Party The Conference Dinner and Party takes place at the Restaurant Bavarie at BMW Welt (Am Olympiapark 1). The dinner starts at 7:30 pm. Public transport: Coming from the station “Max-Weber-Platz”, likewise take the underground U4/U5 (direction “Westend”/”Laimer Platz”). Change at “Odeonsplatz” into U3 (direction “Olympia Einkaufszentrum” to “Olympiazentrum”. From there is is only a few more steps to the BMW Welt. Sponsors We are grateful for the support of our partner and sponsors. 11 Program / Abstracts 12 PROGRAM OVERVIEW Wednesday: April 6th 01:30–02:00 pm Welcome Addresses 02:00–04:00 pm Image-based Systems Biology 04:00 pm Coffee Break 04:30–05:30 pm MTZ Award 05:30–06:15 pm Plenary Talk 06:30 pm Welcome Reception & Poster Session (even numbers) Thursday: April 7th 09:00–11:00 am Single-Cell Systems Biology (Technology & Methods) 11:00 am Coffee Break 11:15 am–01:00 pm Signaling Modeling 01:00 pm 02:00–03:45 pm Lunch Break Metabolism 03:45 pm Coffee Break 04:15–06:00 pm Multi-Scale Approaches 06:00–07:00 pm Poster Session 07:30 pm Conference Dinner & Party at Restaurant Bavarie at BMW Welt Friday: April 8th 09:00–10:45 am Systems Medicine I & Systems Pharmacology 10:45 am Coffee Break 11:15 am–01.00 pm Systems Medicine II & Genetic/Epigenetic Mechanisms 01:00 pm Poster Prize Award & Closing Statement 13 Program / Abstracts Program Wednesday: April 6th 01:30–02:00 pm Welcome Addresses Fabian Theis (Helmholtz Zentrum München) Alfons Enhsen (Helmholtz Zentrum München) Klaus-Peter Michel (Projektträger Jülich, Forschungszentrum Jülich) Peter Jansen (LySiM Network) 02:00–03:45 pm Image-based Systems Biology Chairs: Ivo Sbalzarini & Nassir Navab 02:00 pm Building models of cell organization, differentiation and perturbation directly from microscope images Bob Murphy (Carnegie Mellon University, Pittsburgh) 02:30 pm Image-driven modeling of limb development across 3D space and time Javier Diego (EMBL-CRG, Barcelona) 03:00 pmA quantitative model for macrophage activation predicts tissue thresholds for the propagation of inflammation James Bagnall (University of Manchester, Manchester) 03:15 pm Stem cells dynamics and its regulation during spinal cord regeneration Fabian Rost (Technische Universität Dresden, Dresden) 03:30 pm Tuning out cell cycle entry control – cell fate decisions under oncogenic MYCN Erika Kuchen (DKFZ, Heidelberg) 03:45 pm In silico synchronization and modulation of nuclear ruptures in laminopathy model cells Winnok de Voss (University of Antwerp, Antwerpen) 04:00 pm 04:30–05:30 pm MTZ Award 04:30 pm Introduction and Laudatio Monika Zimmermann, Thomas Zimmermann, Fabian Theis 04:45 pm Robustness of MAPK Signalling Franziska Witzel (Charité Universitätsmedizin Berlin, Berlin) 05:00 pm MicroRNAs decrease protein expression noise Jörn Schmiedel (Charité Universitätsmedizin Berlin, Berlin) 05:15 pm Models of Influenza A Virus Infection: From Intracellular Replication to Virus Growth in Cell Populations Stefan Heldt (University of Oxford, Oxford) 05:30–06:15 pm Plenary Talk Systems Biology in Action: Design of Cancer Combination Therapy Chris Sander (DFCI & HMS, Boston) 06:15 pm Publishing a Paper: the editorial process at Molecular Systems Biology Maria Polychronidion (Molecular Systems Biology) 06:30 pm 14 Coffee Break Welcome Reception & Poster Session (even numbers) Thursday: April 7th 09:00–11:00 am Single-Cell Systems Biology (Technology & Methods) Chairs: Nachiket Vartak & Fabian Theis 09:00 am Comprehensive analysis of single cell states through time and space by mass cytometry Bernd Bodenmiller (University of Zürich, Zürich) 09:30 am Using single-cell genomics to study early development John Marioni (EBI, Hinxton) 10:00 am Single Molecule Approaches For Studying Liver Hetereogeneity Shalev Itzkovitz (Weizmann Institute of Science, Rehovot) 10:15 am Inferring Tumour Evolution from Single-Cell Sequencing Data Edith Ross (University of Cambridge, Cambridge) 10:30 am Comparative analysis of single-cell RNA-sequencing methods Wolfgang Enard (LMU, Munich) 10:45 am Perturbation of dynamics of NF- B and the regulation of gene expression Polly Downton (University of Manchester, Manchester) 11:00 am Coffee Break 11:15 am–01:00 pm Signaling Modeling Chairs: Ursula Klingmüller & Jens Timmer 11:15 am Specificity and Noise in NF- B Signaling Alexander Hoffmann (UCSD, Los Angeles) 11:45 am Signals, effectors, information and decisions David Rand (University of Warwick, Coventry) 12:15 pm Analysing the interplay of NF- B precursors using a mathematical modelling approach Bente Kofahl (Max Delbrueck Center for Molecular Medicine, Berlin) 12:30 pm Identification of an intracellular negative and an autocrine positive feedback coordinating the interferon-induced antiviral response Marcel Schilling (DKFZ, Heidelberg) 12:45 pm Detecting emergent properties in genomic data: Consolidating inflammatory response dynamics Jason Shoemaker (University of Pittsburgh, Pittsburgh) 01:00 pm Lunch Break 02:00–03:45 pm Metabolism Chairs: Madlen Matz-Soja & Hergo Holzhütter 02:00 pm Hepatic ethanol metabolism Ina Bergheim (Freidrich-Schiller-Universität Jena, Jena) 02:30 pm Computational modeling of human metabolism and its application to Parkinson’s disease Ines Thiele (Université du Luxembourg, Luxembourg) 03:00 pm SteatoNet as a predictive and gender-based liver metabolic model Tanja Cvitanović (University of Ljubljana, Ljubljana) 03:15 pm Trans-omic reconstruction of insulin signal flow in global phosphorylation and metabolic network Katsuyuki Yugi (University of Tokyo, Tokyo) 03:30 pm Dynamic glycosylation flux analysis Sandro Hutter (ETH Zürich, Zürich) 15 Program / Abstracts 03:45 pm Coffee Break 04:15–06:00 pm Multi-Scale Approaches Chairs: Dirk Drasdo & Jan Hengstler 04:15 pm Modelling of human haematopoietic cell production to optimize supportive cancer treatment Markus Löffler (IMISE, Leipzig) 04:45 pm Exploiting Evolutionary Trade-Offs as a Novel Cancer Therapy Alexander Anderson (H. Lee Moffitt Cancer Center & Research Institute, Tampa) 05:15 pm Clonal competition in the stem cell niche: New insights from 3D computation Torsten Thalheim (Leipzig University, Leipzig) 05:30 pm Heterogeneity and cell fate control in mouse embryonic stem cells Ingmar Glauche (TU Dresden, Dresden) 05:45 pm Multiscale Model of Liver Intoxication after APAP overdose Noemie Boissier (Sorbonne Universités, Paris) 06:00–07:00 pm Poster Session (odd numbers) 07:30 pm 16 Conference Dinner & Party at Restaurant Bavarie at BMW Welt Friday: April 8th 09:00–10:45 am Systems Medicine I & Systems Pharmacology Chairs: Werner Mewes & Frank Lammert 09:00 am Linking Diseases, Drugs and the Druggable Proteome Tudor Oprea (University of New Mexico, New Mexico) 09:30 am Assimilation of Mammalian Cell Biology knowledge into Model Based Drug Development in Pharmaceutical Industry Andrzej M. Kierzek (Simcyp, Sheffield) 10:00 am Think Adjoint – Methods facilitating parameter estimation for genome-scale mechanistic dynamic models Fabian Fröhlich (Helmholtz Zentrum München |TU München, Neuherberg) 10:15 am Development of an integrative model to improve anemia treatment in non-small cell lung carcinoma patients Agustin Rodriguez-Gonzalez (DKFZ | Heidelberg University | DZL, Heidelberg) 10:30 amAn expandable, multi-level, and multi-scale model for drug simulations of weight-loss and type 2 diabetes 10:45 am Gunnar Cedersund (Linköping University, Linköping) Coffee Break 11:15 am–01.00 pm Systems Medicine II & Genetic/Epigenetic Mechanisms Chairs: Thomas Höfer & Heribert Schunkert 11:15 am Systems Genetic Approaches to Coronary Artery Disease - Toward Diagnostics and Therapies of Molecularly Defined Subcategories of Patients Johan Björkegren (Mount Sinai Hospital, New York) 11:45 am Dissecting analogue versus digital regulation in Polycomb-based epigenetics Martin Howard (John Innes Centre, Norwich) 12:15 pm Model-guided target identification for synergistic combination therapies in the DNA damage response pathway Andreas Raue (Merrimack, Cambridge) 12:30 pm Identification of genetic determinants of immune cellular homeostasis using genetically diverse mouse strains Shai Shen-Orr (Technion-Israel Institute of Technology, Haifa) 12:45 pm Assessing aneuploidy heterogeneity using single cell sequencing Maria Colomé Tatché (Helmholtz Zentrum München, Neuherberg) 01:00 pm Poster Prize Award & Closing Statement 17 Program / Abstracts POSTER OVERVIEW Image-based Systems Biology PP 1-01 Label-free cell cycle analysis for high-throughput imaging flow cytometry Thomas Blasi (Neuherberg) PP 1-02 Cells commit to the cell cycle by rapid and irreversible inactivation of APCCdh1 Steven Cappell (Stanford) PP 1-03 Image-based systems medicine: from mechanistic models to decision-support systems in the clinic Gunnar Cedersund (Linköping) PP 1-04 Investigation of Bile canaliculi formation and biliary transport in 3D in vitro liver cultures Georg Damm (Berlin) PP 1-06 Functional intravital imaging of acetaminophen induced liver injury and regeneration Ahmed Ghallab (Dortmund) PP 1-07 A Systems Survey of Progressive Host Cell Reorganization During Rotavirus Infection Victoria Green (Zurich) PP 1-08 A Qualitative Method of Identifying Stiff Regions in Fibrous Proteins Sean Horan (Irvine) PP 1-10 Image-based modeling of organogenesis Christine Lang (Basel) PP 1-11 Clustering-based classification of autophagy phenotypes in single cell images as a novel readout of autophagic activity Paula Andrea, Marin Zapata (Heidelberg) PP 1-12 Maldi imaging of a prostate cancer biopsy - a pilot study Kjersti Rise (Trondheim) PP 1-13 Complete evaluation of dynamics in an unlabelled live cell from a z-stack of super-resolved bright-field microscopic images Renata Rychtarikova (Nove Hrady) PP 1-14 3D reconstruction and quantitative analysis of liver human samples Fabian Segovia-Miranda (Dresden) PP 1-15 Quantitative analysis of tumor cell load based on diffusion-weighted mri and histology data Yi Yin (Paris) PP 1-16 Investigating Microenvironment-to-cell Signaling in 3D Spheroids through Imaging Mass Cytometry Vito Zanotelli (Zurich) PP 1-17 Analysing the Impact of Errors in Single Cell Tracking Experiments Thomas Zerjatke (Dresden) Single-Cell Systems Biology PP 2-01 destiny – diffusion maps for large-scale single-cell data in R Philipp Angerer (Neuherberg) PP 2-02 Sensitive Detection of Rare Disease-Associated Cell Subsets via Representation Learning Eirini Arvaniti (Zurich) PP 2-03 Information theoretic analysis of interleukin-6-induced signalling by multi-colour flow cytometry Ulrike Billing (Magdeburg) PP 2-04 Refractory states imprinted in the NF- B system regulate encoding of temporal inflammatory signals Christopher Boddington (Manchester) PP 2-05 Wnt/Planar cell polarity signaling regulates commitment of intestinal stem cells to the secretory lineage Anika Böttcher (Garching-Hochbrück) PP 2-06 T cell immune responses generate diversity through linear cell-fate progression 18 Michael Flossdorf (Heidelberg) PP 2-07 Estimating single-cell regulatory heterogeneities from cell populations Christiane Fuchs (Neuherberg) PP 2-08 Universal time and diffusion pseudotime for single-cell data Laleh Haghverdi (Munich) PP 2-09 Dynamical Modeling of Clonal Evolution in Primary and Relapsed Follicular Lymphoma Matthias Horn (Leipzig) PP 2-10 An Interplay of Determinism and Stochasticity in the Information Flow via STAT3 Xiaoyun Huang (Heidelberg) PP 2-11 Dynamic NF- B and E2F interactions control the priority and timing of inflammatory signalling and cell proliferation Nicholas Jones (Manchester) PP 2-12 Inferring cell ensemble models of heterogeneous cell populations by multi-experiment and multi-data-type fitting Stefan Kallenberger (Heidelberg) PP 2-13 Identifying mediators of disease co-morbidities by integrating omics data. Gabi Kastenmüller (Munich) PP 2-14 Unraveling signaling dynamic patterns with single cell mass cytometry Sunil Kumar (Rueschlikon) PP 2-15 Too Young to Die: Age Structured Population Models Capture Cell Cycle Dependent Apoptosis from Snapshot Data Karsten Kuritz (Stuttgart) PP 2-16 GenSSI: Generating Series Structural Identifiability on new Matlab versions Thomas Ligon (Munich) PP 2-17 Mechanistic modeling of subpopulation structures for multivariate single-cell data Carolin Loos (Neuherberg) PP 2-18 Single cell clone transcriptomics derived from murine brown adipose tissue discloses cell type heterogeneity of brown preadipocytes Dominik Lutter (Neuherberg) PP 2-19 Heterogeneity of TLR4 signalling and robust pathogen sensing Pawel Paszek (Manchester) PP 2-20 Cell-to-cell heterogeneity unraveled by computational analysis of single-cell mass cytometry data: cell cycle patterns and trajectories Maria Anna Rapsomaniki (Rüschlikon) PP 2-21 Multiplexed Imaging Cytometry Analysis Toolbox (miCAT) Coupled to Imaging Mass Cytometry (IMC) Reveals Patterns of Cell Interactions Amongst the Heterogeneity of Breast Cancer Denis Schapiro (Zurich) PP 2-22 Inferring gene regulation using pseudotemporal ordering of single cell snapshots F. Alex Wolf (Neuherberg) PP 2-25 Dimensionality Reduction to Networks: Single Cell Sign Consistency Models Enable Identification of Subpopulations with Distinct Signaling Network States in Mammalian Cancer Cells Sophie Tritschler (Zurich) Signalling Modelling PP 3-01 Predicting T-helper cell differentiation and plasticity using logical modeling and model-checking Wassim Abou-Jaoudé (Paris) PP 3-02 Protein abundance of AKT and ERK pathway components governs cell-type-specific regulation of proliferation Lorenz Adlung (Heidelberg) PP 3-03 A system based time series analysis unravels proliferation to differentiation switch in erythroid progenitors cells upon erythropoietin stimulation Geoffroy Andrieux (Freiburg) PP 3-04 The sensitivity of oscillatory properties Katharina Baum (Berlin) 19 Program / Abstracts PP 3-05 Mathematical Modelling Suggests Differential Impact of β-TrCP Paralogues on Wnt/β-Catenin Signalling Dynamics Uwe Benary (Berlin) PP 3-06 Bringing Systems Biology to Material Science: Construction of an Information Processing Material Raphael Engesser (Freiburg) PP 3-07 Mathematical modelling of drug-induced receptor internalisation in HER2-positive breast cancer cell-lines Mirjam Fehling-Kaschek (Freiburg) PP 3-08 Simulation-based parameter estimation for kinetic data for the Raf-Mek-Erk pathway Anna Fiedler (Neuherberg) PP 3-09 Signal transduction analysis reveals key switches in inflammatory signaling during acute myocardial infarction determining disease recurrence Heinrich Huber (Leuven) PP 3-10 Investigating and Modulating Apoptosis Sensitisation in Cultured Cardiac Cells after Exposure to Doxorubicin Using Quantitative Biochemistry and Computational Systems Biology Heinrich Huber (Leuven) PP 3-11 Linking mechanistic understanding of cellular signalling in gastric cancer and cellular phenotypes Sabine Hug (Neuherberg) PP 3-12 Dynamic pathway modeling of IL-6 signaling to unravel mechanisms of Drug-Induced Liver Injury Anja Jünger (Heidelberg) PP 3-13 Auto-correlation of high-precision NF- B oscillation data for dynamic mean population models of TNFα signaling Daniel Kaschek (Freiburg) PP 3-14 Case Study for Attractor Detection of Asynchronous Logical Networks Hannes Klarner (Berlin) PP 3-15 Towards genome-scale mechanistic models of signal transduction Marcus Krantz (Berlin) PP 3-16 Therapeutic Response Modeling Towards Personalized Medicine Jian Li (Munich) PP 3-17 Analysis of the TRAIL and cisplatin induced apoptosis and MAPK-PI3K/AKT signal transduction pathways in melanoma with a probabilistic Boolean network approach Philippe Lucarelli (Belvaux ) PP 3-18 Post-transcriptional regulation by sRNA in Synechocystis PCC 6803 Wolfgang Mader (Freiburg) PP 3-19 A new logic of unified classification of intracellular processes Ozar Mintser (Kyiv) PP 3-20 Dynamic pathway modelling of TNFα signalling to unravel mechanisms of Drug Induced Liver Injury Angela Oppelt (Heidelberg) PP 3-21 Autocrine TGF-beta/ZEB/microRNA-200 signal transduction drives epithelial-mesenchymal transition: Kinetic models predict minimal drug dose to inhibit metastasis Katja Rateitschak (Greifswald) PP 3-22 Tailored Steady-State Constraints for Parameter Estimation in Non-linear Ordinary Differential Equation Models Marcus Rosenblatt (Freiburg) PP 3-23 Dynamics of IL-6-induced classic and trans-signaling Heike Rummel (Magdeburg) PP 3-24 Selective Control of Upregulated and Downregulated Genes by Temporal Patterns and Doses of Insulin Takanori Sano (Tokyo ) PP 3-25 Sensing and antiviral signaling by RIG-I Darius Schweinoch (Greifswald) PP 3-27 One model to rule them all 20 Bernhard Steiert (Freiburg) PP 3-28 Systematic Analysis of Time-Resolved Transcriptional Signature of the Cross-Talk Between HGF and IL6 Reveals Genetic Program of Hepatocyte Proliferation Control Sebastian Vlaic (Jena) PP 3-29 Mathematical modeling of the impact of acetaminophen on the HGF-induced cellular responses in primary mouse hepatocytes Artyom Vlasov (Heidelberg) PP 3-30 Modeling signaling dynamics with differential equations: are single-cell data and their analysis useful? James Wade (Zurich) Metabolism PP 4-01 New Standard Resources for Systems Biology: BiGG Models and Visual Pathway Editing with Escher Andreas Dräger (Tübingen) PP 4-02 Combination of Mass Spectrometry-Based Proteomics and Mathematical Modelling Predicts Therapy Targets of Liver Cancer Merve Erdem (Aachen) PP 4-03 Glucose is not the only source for lactate production by hepatocellular carcinoma cells Jurgen Haanstra (Amsterdam) PP 4-04 Metabolic profiling of CHO-K1 cells adapted to glutamine-free media Michael Hanscho (Vienna) PP 4-05 The relative importance of kinetic mechanisms and variable enzyme abundances for the regulation of hepatic glucose metabolism - Insights from mathematical modeling Hermann-Georg Holzhütter (Berlin) PP 4-06 Significance test for difference between paired temporal observations Ivan Kondofersky (Neuherberg) PP 4-07 A mathematical model of the bile enterohepatic circulation Krystian Kubica (Wroclaw) PP 4-08 New insights in the relation of liver and adipose tissue via Hedgehog Signalling Madlen Matz-Soja (Leipzig) PP 4-09 Why Respirofermentation? Explaining the Warburg effect in tumour (and other) cells by a minimal model Philip Möller (Jena) PP 4-10 Context-specific metabolic modelling reveals cell-type specific epigenetic control points of the macrophage metabolic networkMaria Pires Pires Pacheco (Esch-sur-aLzette) PP 4-11 How Hedgehog signaling pathway activity controls steroidogenesis in the liver Christiane Rennert (Leipzig) PP 4-12 Cause and Cure of Sloppiness in Ordinary Differential Equation Models Christian Tönsing (Freiburg) PP 4-13 Sexual dimorphism during development of cyp51 liver conditional knockout mice Ziga Urlep (Ljubljana) 21 Program / Abstracts Multi-Scale Approaches PP 5-01 Tissue architecture representation in pharmacological models. Insights from liver. Noemie Boissier (Paris) PP 5-02 Agent-based modelling characterises the effect of localized versus spread damage among mitochondrial population Giovanni Dalmasso (Heidelberg) PP 5-03 Morpheus 2.0: an open-source framework for multi-scale multicellular systems biology Walter de Back (Dresden) PP 5-04 Modelling And Simulation Of Tumour Growth Compared With Xenograft Models And Db-scTRAIL Therapy In 2D Cell Culture Population. Simona Galliani (Stuttgart) PP 5-05Data Needs Structure: Data and Model Management for Distributed Research Networks in Systems Biology and Systems Medicine Martin Golebiewski (Heidelberg) PP 5-06 An integrated temporal molecular response of vascular endothelial cells exposed to ionizing radiation Olivier Guipaud (Fontenay-Aux-Roses) PP 5-07 Systems analysis of the structural and molecular changes along the dynamics of liver fibrosis development Seddik Hammad (Mannheim) PP 5-08 Multiscale modeling of liver regeneration Stefan Hoehme (Leipzig) PP 5-10 Modeling Tumor-Immune System Interaction Nick Jagiella (Munich) PP 5-11 Can in vivo 31P MRS assay of myocardial PCr/ATP ratio homeostasis test model predictions? Jeroen Jeneson (Amsterdam) PP 5-12 Multiscale modelling of hepatocellular carcinoma and transarterial chemoembolisation therapy Tanvi Joshi (Stuttgart) PP 5-13 Personalized liver function tests: A Multiscale Computational Model Predicts Individual Human Liver Function From Single-Cell Metabolism Matthias König (Berlin) PP 5-14 Model-based predictions of inflammatory patterns in the breast lobular epithelium in relation to epithelial cell turnover Juan Carlos, Lopez Alfonso (Dresden) PP 5-15 Modeling disease Progression in Myeloproliferative Neoplasms, a Systems Medicine Approach Maryam Montazeri (Aachen) PP 5-16 Development of a Multiscale Systems Biology Approach to Study Atherosclerotic Plaque Progression in WTLdlrKO and M-S196ALdlrKO mice: Integrating Mathematical and Murine Models Cesar Pichardo-Almarza (London) PP 5-17 3D multiscale modeling of vascularized tumor development inside colon Elena Shchekinova (Stuttgart) PP 5-18 Interplay of nucleosome positioning, covalent modifications and transcription factor binding Vladimir Teif (Colchester ) PP 5-19 Modelling of tertiary lymphoid organ development in the context of kidney transplant. Alexey Uvarovskii (Braunschweig) PP 5-20 Parallel analysis of the proteome, phosphoproteome and N-terminome to characterize altered platelet functions in the human Scott syndrome 22 René Zahedi (Dortmund) Systems Medicine & Systems Pharmacology PP 6-01 SIMPLEX: a combinatorial multimolecular omics approach for systems biology Robert Ahrends (Dortmund) PP 6-02 Baseline Matters: the Effect of Initial Immune State on Outcome of Septic Patients Ayellet Alpert (Haifa) PP 6-03 Dynamics of the tumor-infiltrating lymphocyte repertoire in melanoma and pancreatic cancer Lena Maetani Appel (Heidelberg) PP 6-04 Markov-Chain Monte-Carlo methods to analyze mechanistic disease simulators: Applicability and shortcommings Benjamin Ballnus (Neuherberg) PP 6-05 Modeling biliary fluid dynamics reveals possible mechanism for dose-response and personalization of UDCA treatment in PSC Lutz Brusch (Dresden) PP 6-06 Improving non-invasive liver function diagnostics Sascha Bulik (Berlin) PP 6-07 Integrative analysis of chemical high-throughput screens uncovers novel biological information. Monica Campillos (Munich) PP 6-08 A NAT2 Pharmacogenomic based PBPK model of Isoniazid in Men and Its Application in Adjusting Tuberculosis Chemotherapy Henrik Cordes (Aachen) PP 6-09 Validation of a pregnancy population physiologically-based pharmacokinetic model for renally cleared drugs Andre Dallmann (Münster) PP 6-10 A hypothesis-free, systemic approach to identify pharmacological targets for anxiety disorders Michaela Filiou (Munich) PP 6-11 The NormSys registry for modeling standards in systems and synthetic biology Martin Golebiewski (Heidelberg) PP 6-12 Systems biology of MHC class I antigen presentation studied in human cancer cell lines Jennifer Hahlbrock (Mainz) PP 6-13 NormSys & CHARME Two initiatives that aim at Harmonizing the Standardization Processes for Data Exchange in Systems Biology Susanne Hollmann (Potsdam) PP 6-14 A dynamic model of bile acid transport in the HepaRG cell-line Daniel Kaschek (Freiburg) PP 6-15 Modeling Individual Time Courses of Thrombopoiesis During Multi-Cyclic Chemotherapy Yuri Kheifetz (Leipzig) PP 6-16 DESIGNING AND SIMULATION STUDIES OF Mycobacterium Tuberculosis DNA GYRASE. Vidya Niranjan (Bengaluru) PP 6-17 Early biomarkers and magnetic resonance imaging for diagnosing Bronchopulmonary Dysplasia Steffen Sass (Neuherberg) PP 6-18 Using physiologically-based pharmacokinetic modelling to analyze the effect of hepatic impairment on drug detoxification capacity at the whole-body level Arne Schenk (Aachen) PP 6-19 Dynamical modelling of the murine immune response to pneumococcal lung infection with and without antibiotic treatment Sibylle Schirm (Leipzig) PP 6-20 Modelling immuno-chemotherapy of lymphomas Markus Scholz (Leipzig) PP 6-21 Nuclear-encoded mitochondrial genes in hypercholesterolema and atherosclerosis progression Baiba Vilne (Munich) 23 Program / Abstracts Systems Medicine & Genetic/ Epigenetic Mechanisms PP 7-01 Acute and Chronic inflammation and mast cell activation, IN Silico Ablikim Abdukerim (Amsterdam) PP 7-02 Prioritization of candidate causal genes in rare genetic diseases through integration of exome sequencing data and biological databases Ziga Avsec (Garching) PP 7-03 A glance at smoking effect on preeclamptic placenta with transcriptional regulation approach Büşra Aydin (Istanbul) PP 7-04 Post-transcriptional gene regulation in mitochondrial disorder patients Daniel Magnus Bader (Munich) PP 7-05 Quantitative analysis of gene expression based on large distance histone acetylation modifications as enhancers Elham Bavafaye Haghighi (Neuherberg) PP 7-06 Time series analysis of macrophage activation and signalling from RNAseq data Francois Bergey (Berlin) PP 7-07 In situ transcriptomics reveals metabolic and immunological zonation in human liver Mario Brosch (Dresden) PP 7-08 Deep learning of regulatory factors of the transcriptional landscape Gökcen Eraslan (Neuherberg) PP 7-09 Integrant roles of microRNAs and transcription factors in ovarian cancer transcriptional regulatory network Esra Gov (Istanbul) PP 7-10 Cigarette smoking increased expression of the G protein-coupled receptor 15 by change in CpG methylation Tina Haase (Hamburg) PP 7-11 Genetic analysis of spontaneous (non-toxic) liver fibrosis in a congenic mouse model Rabea Hall (Homburg) PP 7-12 ChIP-Seq pipeline for the identification of differential binding of the glucocorticoid receptor in high fat versus low fat diet in mouse livers Johann Hawe (Neuherberg) PP 7-13 Experimentally-based mechanistic modeling to study t helper 17 cell differentiation Jukka Intosalmi (Espoo) PP 7-14 Uncovering potential therapeutic targets and prognostic biomarkers of ovarian tissue related diseases via systems biomedicine approach Medi Kori (Istanbul) PP 7-15 Transcriptional profile of Aldara induced skin-reaction in humans Linda Krause (Neuherberg) PP 7-16 A next generations sequencing toolbox with application to schizophrenia Robert Küffner (Neuherberg) PP 7-17 The properties of Ewing Sarcoma driver cells - widen the view beyond the master fusion gene Nikhil Vinod Mallela (Muenster) PP 7-18 Genome-wide interplay of DNA methylation with genetic variants on the human metabolism Sophie Molnos (Munich) PP 7-19 Reconstruction of the FANCA interactome and discovery of new putative oncogenes in head and neck squamous cell carcinoma. Adriana Pitea (Neuherberg) PP 7-20 Modelling the crosstalk between DNA methylation and histone modifications in cancer Jens Przybilla (Leipzig) PP 7-21 Importance of rare gene copy number alterations for personalized tumor characterization Michael Seifert (Dresden) PP 7-22 Microarrays in large-scale phenotyping at the GMC 24 Julia Söllner (Neuherberg) PP 7-24 ModuleDiscoverer: a novel approach to the discovery of disease modules refining molecular characterization of diet induced rodent models of fatty liver. Christian Tokarski (Jena) PP 7-25 Differential gene expression in atrial fibrillation patients Lars von den Driesch (Neuherberg) PP 7-26 Integrative analysis of exome, transcriptome and cellular bioenergetics profiles in patients with rare mitochondrial diseases Vicente Yepez (Garching) 25 Program / Abstracts The MTZ®-Award for Medical Systems Biology 2016 2016 will be a very important year for the MTZ®foundation. The foundation was established in 2006, so that the MTZ®foundation will celebrates the 10th anniversary. Furthermore it will be the 5th time, that the MTZ®foundation will honour three young scientists for their excellent outstanding dissertations. Adopting the motto “For a better future ...” the MTZ®foundation promotes science and research in the field of Human Medicine. The foundation supports the young generation of scientific talent and promotes their world class and cutting - edge work in biomedical stem cell and genetic research, which is in conformity with the highest bioethical standards. It promotes an interdisciplinary approach, that is, a conscious bringing together of traditional scientific approaches with medical systems biology. Thus, it reaches new dimensions in pathogenic research and drug development. Bioethical questions play a very important role in the work of the MTZ®foundation. Therefore, it is a special concern of her work to indicate the opportunities, but also the limits and risks of this type of research. Bioethical issues play an important role when it comes to maintaining the quality of life within an increasingly ageing society. Germany is a country with much young research talents. The MTZ®foundation aims to support these excellent talents and their potential to be outstanding in the field and seeks to make sure that they get the special attention and public recognition that they deserve. Trial and error and at least success – these elements characterize science and research. These elements are also part of the work of the MTZ-Awardees. The MTZ®foundation wants to motivate them for new research and we are sure, that a MTZ®Award could be a very important step for the further career of the scientific young talents. Scientists from Germany with an international reputation help us to grant awards for research work of world class. The cooperation between the Federal Ministry of Education and Research (BMBF) and these internationally recognized researchers within the MTZ®-Panels provides a rigorous selection of the MTZ®-Awardees. In the meantime the MTZ®-Award is nationally recognized as a national brand - the first important scientific prize for innovative research in the vita of the younger generation of promising scientific talent. Talented young researchers are eager to be discovered and to be promoted early. The MTZ®foundation wishes to be central to that process. That is why it has developed a special students` program to inspire students of the natural sciences and hence, to assure research of world class in Germany in future. A good example is the launch of the virtual MTZ®online “für Schüler”. Participating at this part of the website allows the students of German secondary schools to explore also the fascinating world of medical Systems Biology. 26 MTZ®-Award for Medical Systems Biology 2016 Offering the MTZ®-Award for Medical Systems Biology 2016 should foster future-oriented research approaches in the field of medical Systems Biology. The award is granted by the MTZ®foundation in order to recognize outstanding excellent dissertations in the field of medical Systems Biology, with the aim of attracting special attention and public recognition to these promising researchers. In this endeavour, the MTZ®foundation cooperates with the Federal Ministry of Education and Research (BMBF). The award promotes young scientists that outstandingly contribute to the research in Systems Biology. Our main focus lies on individualised medicine and systems medicine. In this context systems biology is the foundation for a medicine that responds to the individual needs of each patient. The way to a more personalized medicine will become more important in future funding. The prize is divisible and will be assigned for the three best dissertations. The award is being offered for the 5th time. We are very proud to present our three award winners 2016. And we are looking forward with great pleasure to the prize ceremony at the beginning of the SBMC 2016. The managing board Monika Zimmermann Thomas Zimmermann CONTACT For more information about the MTZ®foundation please contact us and visit our website www.mtzstiftung.de. Or just have a look to our new flyer with a short summary of our “MTZ-world” www.mtzstiftung.de/sites/mtzstiftung.de/myzms/content/e191/e705/e6111/ImageFlyer10JahreMTZstiftung.pdf. 27 Program / Abstracts MTZ Award Robustness of MAPK Signalling Witzel F. 1, Fritsche-Guenther, R., Lehmann, N., Sieber, A., Herr, R., Braun, S., Morkel, M., Brummer, T., Sers, C., Blüthgen, N. 1 Charité-Universitätsmedizin Berlin, Germany Variability in gene expression leads to a broad distribution of signalling protein levels in a population of clonal cells, posing a major challenge for the reliability of signal transduction. In my thesis I have investigated how the MAPK pathway, a central signalling pathway controlling various cell fate decisions, remains functional at different expression levels of the terminal kinase Erk1/2. Surprisingly, steady-state signalling remains unaffected when Erk levels are reduced or Erk is highly overexpressed. Moreover, the kinetics and intensity of immediate early gene induction is unaffected when Erk is strongly overexpressed, which indicates the presence of robust pathway features. When analysing kinetic activation data, we find that Erk activity pulses are higher in Erk overexpressing cells when compared to normal Erk levels, but Erk activity is downregulated much quicker. Mathematical modelling points to a combination of fast negative feedback regulation and sequestration phenomena that lead to this strong pathway robustness both at the level of steady-state signaling as well as at pathway output. Taken together, our data show that changes in kinase levels do not necessarily change the pathway activity and might explain the low prevalence of Erk overexpression in diseases such as cancer. MicroRNAs decrease protein expression noise Schmiedel, J. M.1,2,3, Klemm, S.4, Zheng, Y. 3, Sahay, A. 3, Blüthgen, N.1,2,#, Marks, D.S. 5,#, van Oudenaarden, A. 3,6,7,# 1 Institute of Pathology, Charité-Universitätsmedizin, Berlin, Germany 2 Integrative Research Institute for the Life Science and Institute for Theoretical Biology, Humboldt Universität, Berlin, Germany 3 Department of Physics, Massachusetts Institute of Technology, Cambridge, USA 4 Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, USA 5 Department of Systems Biology, Harvard Medical School, Boston, USA 6 Department of Biology, Massachusetts Institute of Technology, Cambridge, USA 7 Hubrecht Institute, Royal Netherlands Academy of Arts and Sciences and University Medical Center Utrecht, Utrecht, The Netherlands #These authors contributed equally MicroRNAs repress many genes in metazoan organisms by accelerating mRNA degradation and inhibiting translation. It has been speculated that microRNAs not only regulate mean protein expression but also protein expression variability, or noise. Here we use mathematical modeling and single cell reporter assays to show that microRNAs – in conjunction with increased transcription - decrease protein expression noise for lowly expressed genes, but increase noise for highly expressed genes. Genes that are regulated by multiple microRNAs show more pronounced noise reduction. We estimate that hundreds of genes in mouse embryonic stem cells have reduced noise due to substantial microRNA regulation. Our findings therefore show how microRNAs confer precision to protein expression and offer plausible explanations for the commonly observed combinatorial targeting of endogenous genes by multiple microRNAs as well as the preferential targeting of lowly expressed genes. Models of Influenza A Virus Infection: From Intracellular Replication to Virus Growth in Cell Populations Heldt, F. S.1, Kupke, S. Y.1, Frensing, T.1,2, Reichl U.1,2 1 2 Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany Chair of Bioprocess Engineering, Otto von Guericke University Magdeburg, Germany Influenza A viruses are human respiratory pathogens that cause annual epidemics with 3 to 5 million cases of severe illness and up to 500,000 deaths. Yet, there are currently only two classes of antiviral drugs licensed for treatment and resistant virus strains are on the rise. In order to elucidate the viral life cycle and identify new targets for antiviral therapy, we constructed a multiscale model of influenza A virus infection. It accounts for the key steps of intracellular virus replication including virus entry, viral genome and mRNA synthesis, and virus release. The model combines this information with infection dynamics at the cell-population level, i.e., the transmission of virions between host cells and the kinetics of virus-induced apoptosis. This integrated modelling approach allows us to capture a variety of experimental data from cell-culture experiments and to predict the most promising targets for direct-acting antivirals. In particular, we find that interference with the viral polymerase represent a highly efficient treatment strategy because it interrupts the autocatalytic cycle of viral RNA synthesis. Moreover, inhibitors of nuclear export and virus assembly/release can readily reduce virus titers. By contrast, targeting the steps of virus entry primarily delays virus spreading but does not protect host cells from infection in vitro. In summary, our multiscale model provides a systems-level understanding of viral infection and is therefore an ideal platform to include further levels of complexity such as the host’s immune response and between-host transmission. 28 INVITED TALKS PT 01 Building models of cell organization, differentiation and perturbation directly from microscope images Murphy, R.1 1 Murphy Lab, Pittsburgh, United States Ray and Stephanie Lane Professor of Computational Biology and Professor of Biological Sciences, Biomedical Engineering and Machine Learning, Carnegie Mellon University, Pittsburgh, USA Honorary Professor of Biology and Senior External Fellow, Albert Ludwig University of Freiburg, Germany Machine learning methods are critically needed for building systems models of cell and tissue behavior and for studying their perturbations. Such models require accurate, cell-type specific information about the shape and distributions of subcellular structures and the distributions of proteins, RNAs and other macromolecules in order to be able to capture and simulate cellular spatiotemporal dynamics. Most efforts towards this goal have used either compartmental models that are not spatially-realistic or geometries constructed from an image of a single cell that are difficult to generate and do not capture cell-to-cell variation. We have developed tools to build generative models of cell organization directly from microscope images of many cells. Generative models are capable of producing new instances of a pattern that are expected to be drawn from the same underlying distribution as those it was trained with. Our open source system, CellOrganizer (http://CellOrganizer. org), currently contains components that can build probabilistic generative models of cell, nuclear and organelle shape, organelle position, and microtubule distribution. These models capture heterogeneity within cell populations, and can be dependent upon each other and can be combined to create new higher level models. The parameters of these models can be used as a highly interpretable basis for analyzing perturbations (e.g., induced by drug addition), and generative models of cell organization can be used as a basis for cell simulations to identify mechanisms underlying cell behavior. Recent work has focused on learning relationships between different cellular structures and structures and on constructing models of the dynamics of changes in cell organization. PT 02 Image-driven modeling of limb development across 3D space and time Diego, J.1 1 EMBL-CRG Systems Biology Research Unit, Centre for Genomic Regulation (CRG), UPF, Barcelona, Spain Organ development is a multi-scale process. At the molecular scale, regulation of genes and proteins creates complex networks which control the activities of cells (cell division, migration, cell fate decisions, differentiation, and many other activities). This molecular control system includes an extracellular part – secreted ligands which move between cells allowing cell-cell communication (such as FGFs, WNTs, etc). The coordination of thousands of cells by this molecular network, leads to large-scale morphogenesis at the scale of tissues and organs. However, these large-scale movements also feedback to the molecular scale: the movement of tissue regions relative to each other, causes cells to receive dynamically changing concentrations of signaling molecules, and this in turn changes the activation or repression of genes. A full understanding of how genes control organogenesis will require multi-scale computer modeling, and we have chosen vertebrate limb development as a model system to explore this problem. Such a model should be based on data about dynamic tissue shapes and spatial distributions of gene activities. Traditional high-throughput and “omics” technologies do not preserve spatial information, and we are therefore using 3D imaging from OPT and SPIM to generate geometric and spatial data for the model. We will present results on our finite-element modeling which is allowing us to tackle this complex problem. PT 03 Systems Biology in Action: Design of Cancer Combination Therapy Sander, C.1 1 Dana-Farber Cancer Institute and Harvard Medical School, Boston, United States Cells and organisms have evolved as robust to external perturbations and adaptable to changing conditions. This capacity poses severe problems for cancer patients. Some targeted anti-cancer drugs work remarkably well, yet resistance to therapeutic perturbation is almost certain to emerge. We address three scientific challenges: (1) Discover the adaptation of cell systems in response to drugs and how to block the escape pathways by combinatorial intervention; 29 Program / Abstracts (2) Describe empirically the landscape of oncogenic alterations for improved and personalized therapeutic approaches, using The Cancer Genome Atlas, and (3) Derive executable network models for cancer cells that guide the development of combination therapy, using perturbation biology (systematic perturbation coupled with rich observation of response, such as changes in protein levels and protein modifications). —— Work done in collaboration with Anil Korkut, Evan Molinelli, Martin Miller, Wei Qing Wang, Xiaohong Jing, Alex Root, Deb Bemis, David Solit, Christine Pratilas, Emek Demir, Arman Aksoy, Onur Sumer, Özgün Babur, Andrea Pagnani, Riccardo Zecchina, Giovanni Ciriello, Nikolaus Schultz, Sven Nelander, Debora Marks et alii. Link: http://bit.ly/1loSE84 - Perturbation biology nominates upstream-downstream drug combinations in RAF inhibitor resistant melanoma cells. PT 04 Comprehensive analysis of single cell states through time and space by mass cytometry Bodenmiller, B.1 1 Bodenmiller Lab, Zurich, Switzerland Each year approximately 1.1 million new cases of breast cancer are diagnosed and about 0.3 million women worldwide die from this disease. Tumor metastases, relapse, and resistance to therapy are the main causes of death in breast cancer patients. Communication between heterogeneous cancer cells and normal cells in the so-called tumor microenvironments (TME) drives cancer development, metastasis formation, and drug resistance. To understand the TME and its relationship to clinical data, comprehensive investigation of the components of the microenvironment and their relationships is necessary. We recently developed a novel imaging modality based on mass cytometry, called imaging mass cytometry (IMC) that enables this type of study. In IMC, tissues are labelled with antibodies that carry pure metal isotopes as reporters. Theoretically 135 antibodies can be visualized simultaneously; in practice, we routinely quantify 44 markers. The antibody abundance is determined using laser ablation coupled to inductively coupled plasma mass spectrometry. Here we demonstrate the capability of IMC and show the results of the analysis of hundreds of breast cancer samples by IMC. To extract biological meaningful data and potential biomarkers from this dataset, we developed a novel computational pipeline for the interactive and automated analysis of large-scale, highly multiplexed tissue image datasets. Our IM data and downstream data analysis revealed a surprising level of inter- and intra-tumor heterogeneity and diversity within known human breast cancer subtypes as well as the stromal cell types in the TME. Furthermore, we identified cell-cell interaction motifs in the tumor microenvironment that correlated with clinical outcomes. In summary, our results show that IMC provides targeted, high-dimensional, subcellular resolved images of tissue samples. The identified spatial relationships among complex cellular assemblies have potential as biomarkers. We envision that IMC will enable a systems biology approach to diagnosis of disease and will ultimately guide treatment decisions. PT 05 Using single-cell genomics to study early development Marioni, J.1 The European Bioinformatics Institute (EBI), Hinxton, Great Britain 1 Gastrulation and the specification of the three germ layers are key events in animal development. However, molecular analyses of these processes have been limited due to the small number of cells present in gastrulating embryos. With recent developments in the field of single-cell biology however, it is now possible to overcome these limitations and to characterize, for the first time at the single-cell level, how cell fate decisions are made. In this presentation I will discuss data generated to study cell fate specification in mouse, as well as the computational strategies we have developed to model such data. I will then illustrate how these data can provide insight into germ layer specification and early erythropoiesis. PT 06 Specificity and Noise in NFkB Signaling Hoffmann, A.1 University of California, Los Angeles, Signaling Systems Lab, Los Angeles, United States 1 Now classic studies revealed complex, stimulus-specific dynamics of NFkB signaling. This led to the hypothesis that the dynamics of signaling constitute a code that allows the cells to transmit information about the external environment (pathogen threats, inflammatory cytokines etc.) to the nucleus to regulate gene expression. However, the information 30 carrying capacity of this dynamical code has remained unclear, largely because available experimental set ups involved cell lines whose adaptation to cell culture conditions rendered them less responsive to many stimuli than primary cells. To examine the responses of primary macrophages, we developed a new NFkB fluorescent reporter mouse and image analysis tools for quantifying NFkB translocation in live cell tracking experiments. These tools enabled unprecedented high throughput experimentation of NFkB activation dynamics allowing us to quantify information transmission capacity and to identify informative dynamical features. These features are recapitulated by surprisingly simple design principles common to ligand-receptor signaling models that are coupled to the core IkB-NFkB model that is capable of producing both oscillatory and non-oscillatory NFkB responses. I will discuss our recent and ongoing studies. Werner, S.L., Barken, D., Hoffmann, A. 2005 Stimulus-specificity of gene expression programs determined by temporal control of IKK activity. Science, 309, pp.1857-61. PMID: 16166517 Behar, M., Barken, D., Werner, S.L., Hoffmann, A. 2013 The Dynamics of Signaling as a Pharmacological Target. Cell, 155, pp.448-461. PMID: 24120141, PMC3856316 Selimkhanov, S.*, Taylor, B.*, Yao, J., Pilko, A., Albeck, J., Hoffmann, A.+, Tsimring, L.+, Wollman, R.+ 2014 Accurate information transmission through dynamic biochemical signaling networks. Science, 346, pp.1270-3. PMID: 25504722. Cheng, Z.*, Taylor, B.*, Rios, D., Hoffmann, A. 2015 Distinct Single Cell Signaling Characteristics Conferred by the MyD88 and TRIF Pathways in TLR4 Activation. Science Signaling, 4(161):ra11 PT 07 Signals, effectors, information and decisions Rand, D.1 University of Warwick, Warwick Systems Biology, Coventry, Great Britain 1 I will discuss a new theoretical approach to information and decisions in signalling systems. In this approach the value of the information in the signalling system is defined by how well it can be used to make the “correct decisions” when those “decisions” are made by molecular networks. As part of this I will introduce a new mathematical method for the analysis and simulation of large stochastic non-linear oscillating systems. This allows an analytic analysis of the stochastic relationship between input and response and shows that for tightly-coupled systems like those based on current models for signalling and clocks, this relationship is highly constrained and non-generic. PT 08 Hepatic ethanol metabolism Bergheim, I.1 1 Friedrich-Schiller-Universität Jena, Biologisch-Pharmazeutischen Fakultät, Jena, Germany Ethanol has been consumed for many centuries and is still one among the leading causes of death and injury world-wide. However, besides being ingested mainly through the consumption of alcoholic beverages ethanol is also constantly produced by intestinal bacteria. Both, ethanol stemming from foods and beverages but also ethanol produced by intestinal bacteria, is due to the lipophilic and hydrophilic properties of ethanol readily taken up in the stomach and even more so in the small intestine. In the liver, alcohol is primarily metabolized by cytosolic alcohol dehydrogenase (ADH) to acetaldehyde which is further metabolized to acetate by mitochondrial aldehyde dehydrogenase (ALDH) 2. Both enzymes use NAD+ as a cofactor, producing a reducing equivalent NADH in both steps. The increased production of NADH has been suggested to be involved in many dehydrogenase-related reactions in the cytoplasm and mitochondria thereby altering energy supply and fatty acids oxidation. Besides ADH, members of the cytochrome P450 family and herein especially the ethanol-inducible cytochrome P 450 2E1 (CYP2E1) are key enzymes involved in the elimination of ethanol, metabolizing ethanol to acetaldehyde using NADPH + H+ as cofactor and molecular O2. Here, ethanol metabolism and its consequences in regards to the damaging effects of ethanol as well as possible modulations of enzymes involved in ethanol metabolism in the liver will be reviewed in detail. PT 09 Computational modeling of human metabolism and its application to Parkinson’s disease Thiele, I.1 Université du Luxembourg, Luxembourg Centre for Systems Biomedicine, Esch-sur-Alzette, 1 The human metabolic network, Recon 2, captures physiological and biochemical knowledge on human metabolism in a stoichiometrically accurate manner and is amenable to computational modeling. Recon 2 has be applied by the systems biology community to understand normal and cellular disease states as metabolism has been shown to play an important 31 Program / Abstracts role in many complex human diseases. I will present our work on using Recon 2 to investigate drug-diet interactions in Parkinson’s disease. The gastrointestinal tract is the site of absorption and metabolism of nutrients and xenobiotics. Several drug-diet interactions take place in the small intestine, such as levodopa (L-DOPA) and high protein diet. These interactions can reduce the drug efficacy. Levodopa has been the gold standard treatment for Parkinson‘s disease treatment since almost 50 years. Despite its efficacy, levodopa leads to fluctuations in the motor response as a consequence of several factors, including diet. Recently, the intestinal transporters of levodopa has been identified in vitro as a neutral and dibasic amino acids exchanger, which provided a molecular basis for the observed competition and suggested a basolateral trans-stimulation of levodopa secretion with amino acids. We have translated these in vitro observations to predict the in vivo bioavailability of levodopa with different diets. Using a combined physiologically based pharmacokinetic and constraint-based reconstruction and analysis approach, we predicted that a serine rich diet could lead to a 10% increase in levodopa bioavailability. A clinical trial published in January 2016 confirmed our predictions. The results showcase the importance of mathematical modeling as a valuable tool for in vitro in vivo extrapolation. Recon 2, and its continuous updates, are accessible through the Virtual Metabolic Human web portal (http://vmh.uni.lu), which links human metabolism to gut microbial metabolism and hundreds of diseases. PT 10 Modelling of human haematopoietic cell production to optimize supportive cancer treatment Markus Löffler (Universität Leipzig, Leipzig) PT 11 Exploiting Evolutionary Trade-Offs as a Novel Cancer Therapy Anderson, A.1 1 Anderson Lab, Tampa, United States Solid tumours export metabolically derived acid into surrounding stroma. We view this microenvironmental acidosis as a niche construction evolutionary strategy in which acid-producing/acid-adapted cancer cell phenotypes benefit by decreasing the fitness of non-adapted stromal competitors, promoting local invasion. These phenotypic properties, in turn, promote transition from in-situ to invasive cancer and a progressive expansion of primary or metastatic tumours. However, there is a significant cost for maintaining an acid-producing/adapted phenotype due to reduced efficiency in energy production and increased energy demand for adaptations to the acidic environment. We hypothesize that this cost may be cancers Achilles’ heel and a novel route for therapeutic intervention. We have generated a multiscale mathematical model that predicts that even subtle perturbations in tumour extracellular pH (pHe) can dramatically alter the emergence and progression of invasive and non-invasive tumour populations. We investigated these eco-evolutionary dynamics and model predictions in the TRAMP (TRansgenic Adenocarcinoma of the Mouse Prostate) model. Consistent with the model predictions, our experiments demonstrated that an increase of intratumoural pHe by only 0.2 pH units prevented transition from in situ to invasive cancer during carcinogenesis and significantly reduced growth and invasion in primary and metastatic cancers. Taken together, our results demonstrate a novel strategy that exploits evolutionary trade-offs to steer the tumour population towards less invasive phenotypes. PT 12 Linking Diseases, Drugs and the Druggable Proteome Oprea T.1 UNM Department of Internal Medicine, Translational Informatics Division, Albuquerque, United States 1 The Illuminating the Druggable Genome Knowledge Management Center (IDG KMC) evaluates, organizes and distils more than 80 protein-centric and over 20 gene-centric resources for over 20,000 curated human proteins (UniProt). The IDG KMC currently focuses on four protein families: G-Protein Coupled Receptors, Nuclear Receptors, Ion Channels and Kinases. However, data integration concerns all human proteins curated in UniProt. Data wrangling, coupled with algorithmic processing, text mining of drug labels, patents, medical literature, as well as human curation and drug-target ontology development, yield emergent properties and knowledge for target-disease associations. Tissue expression data from GTEx, the Human Proteome Map , the Human Protein Atlas and other sources, disease-centric text mining, genomewide association studies and other resources are integrated using a number of specialized ontologies, e.g., the Brenda Tissue Ontology and the Disease Ontology. IDG KMC is also developing a specific drug-target ontology to better integrate 32 chemogenomic and drug information. Using metrics derived from text mining and gene reference into function, as well as the number of antibodies, IDG KMC catalogs 20,204 proteins. Of these, ~38% (7,681) are categorized as “Tdark”, i.e., proteins that lack functional information and disease relevance – the so-called ‘ignorome’. Another 592 proteins (2.9%) have a confirmed drug mechanism of action (“Tclin”) - see box plot. The other two categories reflect levels of literature, functional and disease annotations (“Tbio”), and knowledge about (potent) small molecules (“Tchem”), respectively. From genomics to pharmaco- and clinical informatics, this integrated array of data and machine learning models is currently used to prioritize “Tdark” proteins for further experiment via the 7 NIH funded IDG Technical Development grants, the International Mouse Phenotype Consortium, and other collaborations. An interactive web-based tool, TIN-X, explores the relationship between diseases and proteins, and is currently focused on the 4 IDG protein families. The IDG KMC interface portal, Pharos, supports mining and interactive browsing of this multi-dimensional data collection, which provides informative summaries for the broader scientific community. This integrative effort led to the following observations: i) there appears to be a knowledge deficit, i.e., we lack understanding of protein function for 38% of human proteome; ii) less than 3% of the human proteome is therapeutically addressed by drugs; iii) given current understanding of disease (~8,800 disease concepts), as well as all diseases addressed via on-label (~2,000) and off-label (~400) indications, we currently address at most a quarter of all diseases via therapeutic agents. IDG KMC teams: • Translational Informatics Division, University of New Mexico School of Medicine (Albuquerque, NM); • Ma’ayan Laboratory, Icahn School of Medicine at Mount Sinai (New York, NY); • ChEMBL team, EMBL European Bioinformatics Institute (Hinxton, UK); • Jensen Laboratory, Novo Nordisk Foundation Center for Protein Research (Copenhagen, Denmark); • Schurer Laboratory, University of Miami Miller School of Medicine (Miami, FL); • Division of PreClinical Innovation, NIH National Center for Advancing Translational Sciences (Rockville, MD). Websites: pharos.nih.gov (Pharos) and newdrugtargets.org (TIN-X) PT 13 Assimilation of Mammalian Cell Biology knowledge into Model Based Drug Development in Pharmaceutical Industry Kierzek, A. M.1,2 1 2 Simcyp, a Certara Company, Head of Systems Modelling, Sheffield, Great Britain University of Surrey, Visiting Professor of Systems Biology, Guildford, Great Britain The Physiologically Based Pharmacokinetic Modelling of virtual populations (PopPBPK) is a bottom up, mechanistic modelling methodology, which has already made tremendous impact on the Model Based Drug Development in Pharmaceutical Industry. In a number of cases the U. S. Food and Drug Administration (FDA) has accepted PopPBPK simulation as a sole evidence for drug label, thus decreasing a number of clinical trials required. Building on this unprecedented success of mechanistic modelling approach, Simcyp is now developing Quantitative Systems Pharmacology, where the scope of mechanistic modelling will be extended towards molecular processes in the mammalian cell. The approach will be illustrated by case studies, where kinetic models of Nerve Growth Factor pathway and Fatty Acid Amide Hydrolase inhibitors were extended to incorporate physiological and pharmacological parameters. The possible future directions of incorporating individual genome and ~omics data into genome scale mechanistic models of mammalian cells, as well as further integration of these models with PopPBPK will be discussed. 33 Program / Abstracts PT 14 Systems Genetic Approaches to Coronary Artery Disease - Toward Diagnostics and Therapies of Molecularly Defined Subcategories of Patients Björkegren, J.1 1 Mount Sinai Hospital, New York, United States Cardiometabolic diseases place a heavy burden on society, and new diagnostics, therapies, and strategies for early prevention are needed. Genome-wide association studies (GWAS) have identified hundreds of new genetic risk loci, thereby improving our understanding of the complex inheritance of cardiometabolic disease. However, their contribution to expected genetic variance is typically low, and most of the downstream gene-regulatory mechanisms remain unknown. We genotyped and RNA-sequenced seven disease-relevant tissues from 600 coronary artery disease patients in the Stockholm-Tartu Atherosclerosis-Related Network Engineering Task (STARNET) study. The STARNET datasets were highly informative in unfolding mechanisms for risk loci identified by GWAS. We unexpectedly found that the principal site for the genetic regulation of the LDL-cholesterol and CAD risk gene PCSK9 appears to be abdominal fat, not liver. In addition using strict statistical thresholds, we found evidence of extensive sharing of cis and trans genes across tissues and diseases by revealing casual disease networks of multiple loci. A better understanding of downstream gene-regulatory mechanisms of risk loci for cardiometabolic diseases provided by STARNET is essential to translate the initial GWAS findings into opportunities for diagnosis, therapy and prevention. PT 15 Dissecting analogue versus digital regulation in Polycomb-based epigenetics Howard, M.1 1 John Innes Centre, Dept of Computational and Systems Biology, Norwich, Great Britain Polycomb repressive complex 2 (PRC2) is involved in transcriptional repression of thousands of genes in higher eukaryotes. Expression of a PRC2 target gene is not only controlled digitally in cis by its chromatin state, but also in an analog (smoothlyvarying) manner by gene-specific trans-regulators. Here, we introduce a mathematical model linking cis- and transregulation, where transcription directly antagonizes Polycomb silencing. Our model reveals that locally acting feedbacks stabilizing the PRC2-repressed state allow chromatin to buffer noise in trans-regulators. Buffering is particularly robust when chromatin dynamics are slow, suggesting a biological function for observed slow histone methylation by PRC2. When trans inputs to gene expression are balanced, bistable chromatin states instruct their own inheritance, with analog transregulators fine-tuning expression of the two states. However, strong transcriptional activation or repression abolishes bistability, with chromatin becoming purely responsive to trans-factors. Our model therefore clarifies causality relationships between transcription and chromatin in epigenetic memory. 34 ORAL PRESENTATIONS OP 01 A quantitative model for macrophage activation predicts tissue thresholds for the propagation of inflammation Bagnall, J.1, Boddington, C.1, Downton, P.1, Boyd, J.1, Brignall, R.1, Rowe, W.1, Schmidt, L.1, Spiller, D.1, White M.1, Paszek, P.1 University of Manchester, Manchester, Great Britain 1 Toll-like receptor (TLR) signaling regulates macrophage cell activation and effector cytokine amplification in response to pathogen stimulation. Here, using quantitative single cell approaches and mathematical modeling we propose that TLR4 signaling involves a “tissue level feedback”, which regulates propagation of the effector cytokine response by competitive uptake in a local tissue environment. The project was undertaken using a combination of fluorescence correlation spectroscopy calibrated microscopy, single-molecule FISH and flexible recombination lentiviral platform. We found that in macrophage populations, the extent of TLR4 stimulation defines a dose-dependent expression of Nuclear Factor -kappaB (NF- B) target genes. However, only a subset of macrophages was able to produce a strong Tumor Necrosis Factor a (TNFα) response, which resulted in cellular heterogeneity. Based on the measured high TNFα uptake seen in tissue cells (in comparison with macrophages), mathematical modeling suggests that the local propagation of the TLR4 cytokine effector response is limited to small distances of a few cell diameters. In our predictive model, the heterogeneity of TNFα production controlled the requisite interaction distance and the probability of signal propagation between tissue-resident macrophage cells. We propose that “tissue-level feedback” may allow robust amplification of localized inflammatory cues, while avoiding out-of-context propagation of inflammation at long range. OP 02 Stem cells dynamics and its regulation during spinal cord regeneration Rost, F.1, Rodrigo Albors, A. 2,3, Mazurov, V. 2, Deutsch, A.1, Brusch, L.1, Tanaka, E. M. 2, Chara, O.1,4 Technische Universität Dresden, Center for High Performance Computing, Dresden, Argentina Technische Universität Dresden, Center for Regenerative Therapies, Dresden, Germany 3 University of Dundee, Cell and Developmental Biology, Dundee, Great Britain 4 National University of La Plata, Instituto de Física de Líquidos y Sistemas Biológicos, La Plata, Argentina 1 2 Question: Regeneration, the ability to recreate lost tissues is a long studied question in biology. Strikingly, mammals have generally poor regenerative abilities. On the contrary, the Mexican salamander or axolotl is uniquely able to mobilize neural stem cells to completely regenerate many lost neural tissues including the spinal cord. An outstanding biological question is why this animal can and why so many other species, including mammals, can’t. To answer this question we first need to understand what the cellular mechanisms controlling axolotl spinal cord regeneration are. After tail amputation the spinal cord regrows about 2 mm within 1 week. What are the cell scale mechanisms that drive this tissue outgrowth? Methods & Results: Morphogenetic processes like cell proliferation, and (de-)differentiation, cellular rearrangements are obvious candidates. We aimed to quantify them by tightly linking quantitative data analysis and mathematical modelling. We quantified the density of neural stem cells and mitotic events along the anteroposterior axis in the axolotl spinal cord during the first week of regeneration. We analyzed these data with an empirical mathematical model of two spatial zones of cell proliferation along the anteroposterior axis using Bayesian inference. This analysis indicated a region close to the amputation plane that shows an increase in cell proliferation. To estimate the proliferation rate dynamics, we quantified and modelled the incorporation of a thymidine analog (BrdU) in the regenerating spinal cord. Furthermore, we tracked individual cell clones to estimate the rate of cell displacements and hence the flux of cells into the regenerating zone. Finally, we set up a mechanistic mathematical model of spinal cord outgrowth during regeneration that incorporates proliferation, differentiation and cell displacements. By comparing this model to spinal cord outgrowth data we can show that the proliferation in a zone close to the amputation plane is the key to explain the observed outgrowth. 35 Program / Abstracts OP 03 Tuning out cell cycle entry control - cell fate decisions under oncogenic MYCN Kuchen, E.1, Ryl, T.1, Bell, E. 2, Shao, C.1, Florez, A.1, Westermann, F. 2, Höfer, T.1 1 2 DKFZ, Theoretical Systems Biology, Heidelberg, Germany DKFZ, Neuroblastoma Genomics, Heidelberg, Germany While most tumours respond to initial therapy, tumour relapse remains a major medical challenge. Fundamentally, tumour regression or regrowth are determined by the individual decisions of cancer cells under treatment, including cell death, senescence and continued proliferation of survivors. Eliminating viable survivors relies on a mechanistic molecular understanding of the processes that determine the fate of individual cells and their deregulation in cancer cells. Key regulators of all these cellular decisions are the members of the MYC family of transcription factors. MYC is frequently overexpressed in a multitude of human cancers and is one of the most prominent drivers of tumour relapse. MYC’s promiscuous binding abilities make it difficult to understand its role in cellular decision-making. Using cells of the paediatric cancer neuroblastoma harbouring a MYCN amplification, we investigated whether the bistable switch, underlying both the cell cycle G1-S transition and DNA-damage arrest, is still present under oncogenic MYCN. Transcriptomics of cell cycle synchronised populations with tuneable MYCN expression revealed that specific MYCN-induced changes in gene expression shift the balance between positive and negative regulators of this bistable switch. Using a data-driven mathematical model, we demonstrate that, through this shift, the switch is tuned out under high MYCN expression and unperturbed growth, thus pushing cellular decisions towards proliferation. This detuning homogenises cell cycle progression and increases the rate of passage through the G1-phase, as validated by single-cell time-lapse microscopy and flow cytometry experiments. However, even in these aggressive cancer cells, the bistable switch is not abolished and can be restored by pharmacological perturbations. Corroborated by experiments, the model shows that cells remain responsive to chemotherapy-induced cell cycle arrest, coupling DNA-damage and cell cycle checkpoints. Upon chemotherapy, MYCN delays cell cycle arrest and thus shifts treatment effects from arrest to apoptosis. This effect is accentuated after chemotherapy, where high MYCN levels suppress cellular senescence and drive surviving cells back into the cell cycle. Our quantitative analyses provide a mechanistic understanding of how MYCN shifts single-cell decisions from quiescence to proliferation under unperturbed growth and suggest that long-term arrest after chemotherapy is maintained via the same mechanism. OP 04 In silico synchronization and modulation of nuclear ruptures in laminopathy model cells Robijns, J.1, Molenberghs, F.1, Corne, T.1,2, Sieprath, T.1,2, Verschuuren, M.1, De Vos, W.1,2 University of Antwerp, Antwerp, Belgium University of Ghent, Molecular Biotechnology, Ghent, Belgium 1 2 The nuclear lamina is a critical regulator of nuclear structure and function. We have shown that nuclei from laminopathy patient cells experience repetitive disruptions of the nuclear envelope, causing transient intermingling of nuclear and cytoplasmic components (De Vos et al., Human Molecular Genetics, 2011). Ruptures occur at weak spots of the nucleus, pointing to mechanical defects, while the uncontrolled translocation of transcription factors during those events alter gene expression programs. Since ruptures have recently also been described in aging and cancer cells - both associated with abnormal expression of lamins or their precursors - it most likely represents a pathophysiological mechanism with generic relevance. However, the exact causes and consequences of rupture events are not fully understood, and their stochastic occurrence complicates in-depth analyses. To resolve this, we have established a method that enables quantitative investigation of nuclear ruptures, based on minimally invasive imaging of two fluorescent reporters, automated cell tracking and in silico synchronization of individual rupture events. Using this approach, we systematically quantified nuclear rupture kinetics in laminopathy model cell lines. We found that rupture frequency correlates inversely with lamin A/C levels in model cells and we discovered novel regulators of rupture kinetics. OP 05 SINGLE MOLECULE APPROACHES FOR STUDYING LIVER HETEREOGENEITY Itzkovitz S.1 Weizmann Institute of Science, Rehovot, Israel 1 The liver is a highly heterogeneous organ composed of polyplooid hepatocytes operating in structured lobules, which are polarized by blood flow. The polarized blood flow along the lobule axis creates gradients of oxygen, nutrients and hormones. As a result different lobule layers sub-specialize in distinct processes, a phenomenon termed ‘metabolic zonation’. To study liver heterogeneity in the intact mammalian liver we have implemented single molecule Fluorescense in-situ Hybridization (smFISH), a technique that reveals individual mRNA molecules as diffraction limited dots under a fluorescence microscope. This technique facilitates quantification of the zonation profiles of any liver gene of interest. We also developed techniques, 36 based on dual color labeling of introns and exons, to extract both transcription and degradation rates of genes in the intact mouse liver. Using these methods we found that gluconeogenic genes such as Pck1 and G6pc have extremely high transcription and mRNA degradation rates, facilitating rapid shut-down upon refeeding. Many liver genes are expressed in a bursty manner, with promoters stochastically transitioning between active transcription and quiescence. Strikingly the liver seems to have developed features to minimize cell-to-cell variability generated by such bursts, including coordination of burst frequency and mRNA lifetime, liver polyplopidy and nuclear retention of mRNA for highly bursty genes. Our results reveal new layers of liver heterogeneity. OP 06 Inferring Tumour Evolution from Single-Cell Sequencing Data Ross, E.1, Markowetz, F.1 University of Cambridge, Cancer Research UK Cambridge Institute, Cambridge, Great Britain 1 Question: Tumour evolution leads to genetic intra-tumour heterogeneity, which poses major challenges to cancer therapy. While tumour heterogeneity has been documented in several cases, many details of the underlying evolutionary processes are still unknown. Studying pathways of tumour evolution promises to provide insights into these processes and the early stages of cancer development and could enable predictions about whether or not early-stage tumours are likely to progress to more aggressive forms. Inferring the evolutionary history of a tumour, however, remains challenging. So far, most methods for inferring tumour phylogenies use bulk sequencing data, but these methods struggle to deconvolute the mixed signal into separate clones and their corresponding genotypes. Recent advances in single-cell sequencing technologies promise to reveal tumour heterogeneity at a much higher resolution. However, single-cell sequencing data come with their own challenges and new methods are needed that take into account the noise that is inherent to this type of data. Results: Here, we present oncoNEM (oncological Nested Effects Model), a probabilistic method for inferring intra-tumour evolutionary lineage trees from noisy exome- or genome-wide single-cell sequencing data. OncoNEM is based on the nested structure of mutations observed between cells and jointly infers the tree structure, the number of clones and their composition. We evaluate the accuracy of oncoNEM in the controlled setting of simulation studies and demonstrate that (i) our method can accurately infer trees of tumour evolution despite the high allelic dropout rates of current single-cell sequencing technologies, (ii) it can estimate error rates directly from the data and is robust to inaccuracies in these model parameters and (iii) it substantially outperforms competing methods. Finally, we show its applicability in case studies of muscle-invasive bladder cancer and essential thrombocythemia. Conclusions: OncoNEM is an accurate probabilistic method for inferring intra-tumour phylogenies from single-cell sequencing data. It identifies subpopulations within a sample of single-cells and estimates their evolutionary relationships. In simulations, oncoNEM performs well for false-positive and false-negative rates of current single-cell data sets, it is robust to inaccuracies in the estimation of model parameters and outperforms competing methods. 37 Program / Abstracts OP 07 Comparative analysis of single-cell RNA-sequencing methods Enard, W.1, Ziegenhain, C.1, Parekh, S.1, Vieth, B.1, Smets, M.1, Leonhardt, H.1, Hellmann, I.1 1 LMU Munich, Biology II, Planegg, Germany Single-cell mRNA sequencing (scRNA-seq) allows to profile heterogeneous cell populations, offering exciting possibilities to tackle a variety of biological and medical questions. A range of methods has been recently developed, making it necessary to systematically compare their sensitivity, accuracy, precision and cost-efficiency. Here, we have generated and analyzed scRNA-seq data from 479 mouse ES cells and spike-in controls that were prepared with four different methods in two independent replicates each. We compare their sensitivity by the number of detected genes and by the efficiency with which they capture spiked-in mRNAs, their accuracy by correlating spiked-in mRNA concentrations with estimated expression levels, their precision by power simulations and variance decomposition and their efficiency by their costs to reach a given amount of power. While accuracy is similar for all methods, we find that Smart-seq on a microfluidic platform is the most sensitive method, CEL-seq is the most precise method and SCRB-seq and Drop-seq are the most efficient methods. Our analysis provides a solid basis to choose among four available scRNA-seq methods and to benchmark future method development. OP 08 Perturbation of dynamics of NF- B and the regulation of gene expression Downton, P.1, Ashall, L.1, Rowe, W.1, Adamson, A.1, Boddington, C.1, Harper, C.1, Boyd, J.1, Daniels, D.1, Lam, C.1, Ryan, S.1, Spiller, D.1, Paszek, P.1, White, M.1 University of Manchester, Faculty of Life Sciences, Manchester, Great Britain 1 The nuclear factor kappa B (NF- B) signalling system is critically important with respect to inflammation, immune responses and cell survival. NF- B regulates the expression of hundreds of genes and is activated by a large range of biological factors and environmental conditions. Uncontrolled regulation of the NF- B pathway can lead directly to a number of disorders, including chronic inflammation and various cancers. Imaging of single cells reveals that induction of NF- B signaling by tumour necrosis factor alpha (TNFalpha) results in oscillations of NF- B between the cytoplasm and nucleus [Nelson et al., (2004) Science 306:704], where NF- B can bind promoters to induce target gene transcription. Continuous treatment with TNFalpha results in persistent nuclear oscillations that lose synchronicity over time, with an average period of 100 minutes between translocations. When cells are treated with pulses of TNFalpha at 100 minute intervals, the majority of cells show a relatively synchronous nuclear translocation in response to each treatment [Ashall et al., (2009) Science 324:242]. We have modelled this oscillatory process mathematically, and characterized the importance of the molecular feedback mechanisms in heterogeneity in this system [Paszek et al., (2010) Proc. Natl. Acad. Sci. 107:11644]. To determine the functional relevance of these system properties, we conducted a large scale microarray analysis of gene expression in response to continuous and pulsatile TNFalpha treatment regimens. We identify hundreds of induced target genes, and find that different subsets of NF- B-dependent genes respond to continuous and pulsatile stimulation protocols. We have also perturbed both NF- B oscillation dynamics and target gene expression using combinations of physiological stimuli. Target gene differences are visible at a transcript and protein level, suggesting oscillation frequency and synchronicity have functional rolesin the regulation of downstream gene expression. Single molecule fluorescence in situ hybridisation (smFISH) has been used to characterise the heterogeneity of target gene expression at a single cell level, and these data are being used to inform and refine our mathematical models of NF- B-dependent transcription. 38 OP 09 Analysing the interplay of NF- B precursors using a mathematical modelling approach Kofahl, B.1, Yilmaz, Z. B.1, Beaudette, P.1, Baum, K.1, Ipenberg, I.1, Dittmar, G.1, Scheidereit, C. 1, Wolf, J.1 1 Max Delbrueck Center for Molecular Medicine, Berlin, Germany NF- B signalling is involved in inflammatory responses, innate and adaptive immunity, cell proliferation and cell death. Its aberrant regulation is linked to inflammatory and cardiovascular diseases as well as to cancer. Generally, the NF- B signalling network is divided into the canonical and the non-canonical branch. They are activated by different stimuli and act on different time scales. The non-canonical branch relies on the processing of the precursor protein p100 to its product p52. Via the canonical branch, the amount of p50 (generated by the precursor p105) that can enter the nucleus is regulated. In recent years, it has been shown that both branches are interconnected, e.g. via target gene expression. Our combined experimental and theoretical study reveals a further linkage between the two branches by interplay of the two precursors. Our experimental data showed that the dynamics of p100 and p105 proteolysis are similar upon pathway activation by a noncanonical stimulus and that the precursors can form a hetero-complex. Furthermore, the data showed that their dynamics are interdependent - the presence of one precursor influences the temporal behaviour of the respective other one. Based on the experimental findings, different mathematical models are developed and used to dissect the mechanism underlying the concerted response of p100 and p105. Comparing the fit quality of the models reveals that the p100/p105 complex is stimulus-responsive. Quantitative dissection and modeling of the NF- B p100-p105 module reveals interdependent precursor proteolysis. Yilmaz, ZB; Kofahl, B; Beaudette, P; Baum, K; Ipenberg, I; Weih, F; Wolf, J; Dittmar, G; Scheidereit, C Cell Rep. 2014, 9(5): 1756-1769 OP 10 Identification of an intracellular negative and an autocrine positive feedback coordinating the interferon-induced antiviral response Schilling, M.1, Robichon, K.1, Maiwald, T. 2, Schneider, A.1, Willemsen, J. 3, Kreutz, C. 2, Ehlting, C.4, Chakraborty, S.1, Huang, X.1, Böhm, M. E.1, Damm, G. 5, Seehofer, D. 5, Bode, J. G.4, Binder, M. 3, Bartenschlager, R.6, Timmer, J. 2, Klingmüller, U.1 German Cancer Research Center (DKFZ), Division Systems Biology of Signal Transduction, Heidelberg, Germany University of Freiburg, Institute for Physics, Freiburg, Germany 3 German Cancer Research Center (DKFZ), , Division Virus-Associated Carcinogenesis, Heidelberg, Germany 4 University of Düsseldorf, University Hospital, Gastroenterology, Hepatology and Infect. Diseases, Düsseldorf, Germany 5 Charité-University Medicine, General, Visceral, and Transplantation Surgery, Berlin, Germany 6 University of Heidelberg, Department of Infectious Diseases, Heidelberg, Germany 1 2 Interferon alpha (IFNα)-induced activation of an antiviral response is the first line of defense against viral infection. Initiation of an antiviral response comprises the induction of a multitude of antiviral genes that remarkably vary in their extent and dynamics of expression. By combining time-resolved quantitative data and mathematical modeling, we established mRNA stability and a negative as well as a positive feedback loop as key mechanisms controlling the expression dynamics of the antiviral genes. We show that distinct antiviral gene expression profiles are induced by IFNα stimulation. Iterating between experiment and modeling, we analyzed the pathway structure using a dynamic model of IFNα-induced signaling and gene expression. Model prediction and experimental validation revealed that distinct mRNA stability affects expression profiles of antiviral genes and that the intracellular feedback mediated by an interferon-induced gene negatively regulates expression of early antiviral genes. Furthermore, model analyses predicted an autocrine feedback, which cooperates with IFNα to enhance antiviral gene expression. We experimentally identified the key mediator of this positive feedback loop by mass spectrometry and by employing knock-out mice. Furthermore, we developed model-based strategies to potentiate the IFNα-induced antiviral responses. This effect could be therapeutically exploited to promote IFNα-mediated viral clearance and thereby antagonize viral persistence. 39 Program / Abstracts OP 11 Detecting emergent properties in genomic data: Consolidating inflammatory response dynamics Shoemaker, J.1,2, Satoshi, F. 2, Eisfeld, A. 3, Zhao, D. 2, Kawakami, E. 2, Sakabe, S. 2, Maemura, T. 2, Goria, T. 2, Katsura, H. 2, Muramoto, Y. 2, Watanabe, S. 2, Watanabe, T. 2, Fuji, K. 2, Matsuoka, Y. 2, Kitano, H. 2, Kawaoka,Y. 2,3 University of Pittsburgh, Chemical & Petroleum Engineering, Pittsburgh, United States ERATOInfection-induced Host Responses Project, Tokyo, Japan 3 University of Wisconsin, Madison, Japan 1 2 Main Objectives: Understanding the emergent behaviors that result from molecular interactions is essential to detecting divergent outcomes. This is particularly true in biology when a common mechanism may drive distinct dynamic results. Bottom-up modeling approaches that rely on mathematical models can help define the feasible dynamic landscape of the biological system but such approaches are difficult in genomic studies when system size can overwhelm applicability. Here, we describe an approach to leverage bioinformatics and clustering to screen for select emergent behaviors in whole genome expression data. The approach is applied to gene expression data from the lungs of mice infected with influenza viruses to evaluate whether different virus strains invoke unique inflammatory gene responses or if emergent properties, e.g. ultrasensitive responses, can consolidate apparent, virus-specific responses. Materials and Methods: Animal models: Mice were infected with a seasonal H1N1, an H1N1 virus from the 2009 pandemic or a deadly H5N1 virus. At several times post infection, three animals per group were sacrificed and the lungs sectioned for microarray analysis and virus titer assays. Module detection and modeling: Dynamic gene expression was clustered using WGCNA and screened for association with inflammation using a variety of bioinformatics techniques. Inflammation-associated gene modules were fit to equations relating gene response to virus titer. Model validation: To evaluate the fitted models, the animal study was repeated but the inoculation levels altered. The model prediction was evaluated by goodness of fit. Results: We found that a single module of clustered genes was highly enriched for inflammatory signaling. The module’s scaled mean gene expression appeared to have an ultrasensitive-like relationship with virus growth, i.e. minimal gene response is observed at low virus concentration but gene expression increases rapidly once a threshold virus titer was reached. Fitting the gene expression to a simplified model of ultrasensitivity, we showed that not only could the model consolidate the inflammatory gene expression dynamics that appeared distinct to each virus, but we could also predict inflammatory gene expression in newly infected populations. Conclusions: The findings of this work strongly suggest that ultrasensitive responses coordinate the inflammatory response and are responsible for major inflammatory events during virus infections. Furthermore, this study demonstrates the importance of integrating signaling framework in to gene expression analysis. OP 12 SteatoNet as a predictive and gender-based liver metabolic model Cvitanović, T.1, Moškon, M. 2, Mraz, M. 2, Urlep, Z.1, Rozman, D.1 1 2 Institute of Biochemistry, Faculty of Medicine, University of Ljubljana, Center for Functional Genomics and Bio-chips, Ljubljana, Slovenia Faculty of Computer and Information Science, University of Ljubljana, Ljubljana, Slovenia Several large-scale computational models of liver metabolism address the liver dynamics from systems biology/medicine perspective. Herein we describe applications of the state-of-the-art model SteatoNet designed by the object-oriented approach (Naik et al., PLOS Comput. Biol., 2014). SteatoNet accounts for interactions between the liver and peripheral tissues and includes metabolic as well as gene regulatory and signal transduction pathways describing the dynamics of non-alcoholic fatty liver disease. SteatoNet requires only a minimal set of parameters and can be used even in a case of sparse experimental data. Furthermore, due to its object-oriented nature, it can be easily adapted to investigate different liver-associated pathologies. This makes the model an excellent starting point for testing biological hypotheses prior to experimentation. We applied SteatoNet to address the question of metabolic consequences in adipose tissue after knocking out gene Cyp51 from cholesterol synthesis in the liver. The experimental data for the Cyp51 liver knockout are available (Lorbek et al., Sci. Rep. 2015). The model simulations demonstrate the network disturbances in adipose tissue, which is an excellent starting point for further experimental testing on gene expression and protein levels. Another important application is the gender-based model adaptation. Liver has been known for decades as a sexually dimorphic organ especially at the gene expression level. Gender-based differences were discovered also in the experimental Cyp51 liver knockout responses. We extended SteatoNet to differentiate between genders based on literature data and expert based knowledge. As far as we can tell this represents the first gender-based liver metabolic model. Current applications include simulations of sex hormone ratios in blood and their networking with gender-based differences in cholesterol synthesis and regulatory nodes. The future SteatoNet adaptation will be guided towards personalization, aimed at predicting the network effects of liver disease -related polymorphisms in individuals. 40 OP 13 Trans-omic reconstruction of insulin signal flow in global phosphorylation and metabolic network Yugi, K.1, Kubot, H. 2, Kuroda, S.1 University of Tokyo, Department of Biological Sciences, Tokyo, Japan Kyushu University, Medical Institute of Bioregulation, Fukuoka, Japan 1 2 Cellular functions are realized by a global network of molecular interactions across multiple ‘omic’ layers such as genome, transcriptome, proteome, and metabolome. We designate this global network as a ‘trans-omic’ network. Conventional molecular biological studies and comprehensive measurement of each omic layer have collectively provided clues to elucidate the landscape of the trans-omic network. However, conventional molecular biology has limited comprehensiveness, and indepth measurement of one particular omic layer does not reveal interaction across multiple omic layers. Consequently, the landscape of the trans-omic network remains unknown. Here we developed an unbiased method to reconstruct trans-omic networks based on time-series phosphoproteome and metabolome data together with public databases. This method was applied to reconstruct a trans-omic network underlying acute insulin action in rat hepatoma FAO cells. We found that the insulin signal flowed through the trans-omic network involving 26 protein kinases, 76 phosphorylated metabolic enzymes, and 80 allosteric effectors, resulting in quantitative changes in 97 metabolites. Kinetic modeling analysis predicted selective control of a subnetwork around phosphofructokinase by specific phosphorylation and allosteric regulation. Thus, we provide an unbiased method that reconstructs the trans-omic network from phosphoproteome and metabolome data, which will be applicable to other cellular responses. OP 14 Dynamic glycosylation flux analysis Hutter, S.1, Villiger, T. K.1, Brühlmann, D. 2, Stettler, M. 2, Broly, H. 2, Morbidelli, M.1, Gunawan, R.1 1 2 ETH Zürich, Chemie und Angewandte Biowissenschaften, Zürich, Indonesia Merck Serono S.A., Biotech Process Sciences, Corsier-sur-Vevey, Switzerland Objectives: N-linked glycosylation is essential in determining the biological activity and efficacy of many therapeutic proteins [1]. Hence, there is a great interest for modulating and controlling protein glycosylation to optimize and regulate product quality. Much of the previous developments have been driven by the optimization of process parameters and media composition, and by genetic engineering [2]. Engineering approaches that rely on mathematical modeling of glycosylation networks however have not been adopted widely for this purpose. In this work, we present a method for dynamic flux analysis of glycosylation networks from time series measurements of glycan structures, aptly called dynamic glycosylation flux analysis (DGFA) and apply it to a CHO cell culture production of IgG monoclonal antibody. Materials & Methods: In developing DGFA, we adapted a previous method called dynamic metabolic flux analysis (DMFA) [3]. DGFA, like DMFA, relies on a stoichiometric model of the network which can be written as: dX/dt = Nv Sv = 0 X denotes the external concentrations of glycosylated protein products, v the vector of glycosylation fluxes, and N and S respectively denote the external and internal stoichiometric matrices. DGFA generates estimates for the fluxes v at an optimally determined set of switching time points, by minimizing the sum of squares of the errors (SSE) between the measured concentrations and model predictions of X(t). We employed a piecewise linear approximation for v(t) between the switching times. In typical glycosylation networks, the stoichiometric matrix S is underdetermined due to branching and converging pathways. To resolve the ambiguity in flux predictions caused by such pathways, we employed an equal flux ratio between fluxes that produce the same glycosylation structure. In addition, we also perform flux uncertainty analysis by assigning a random ratio to the fluxes above. Glycosylated protein samples in a CHO-S fed-batch cell culture for the production of IgG monoclonal antibody were analyzed by Ultra Performance Liquid Chromatography (UPLC) over a period of 14 days together with measurements of antibody 41 Program / Abstracts concentrations and cell count measurements. Results & Conclusion: DGFA could match the measurement data with excellent accuracy. Statistical analysis resulted in two optimal switching time points on day 7 and 9. A comparative analysis revealed that glycosylation reactions catalyzed by the same enzyme evolved similarly over time and uncertainty analysis showed that DGFA was robust with respect to the flux ratios used in the analysis. Hence, the glycosylation flux predictions from DGFA could reveal the time-evolution of enzyme activities, an important insight that could lead to a better understanding of the regulation of the glycosylation process. [1] Aebi, M., 2013. Biochim. Biophys. Acta. [2] Hang, I., 2015. Glycobiology. [3] Leighty, R.W., 2011. Metab. Eng. OP 15 Clonal competition in the stem cell niche: New insights from 3D computation Thalheim, T.1, Buske, P.1, Rother, K.1, Przybilla, J.1, Loeffler, M.1, Galle, J.1 1 Leipzig University, IZBI, Leipzig, Germany Competition between stem cell (SC) clones represents a mechanism that potentially can eradicate mutant cells and their progeny from a tissue. In the intestine this competition leads to ongoing monoclonal conversion in all individual crypts of the tissue. General features of this process have been studied experimentally. However, the impact of the spatial organization of the intestinal crypts on the fixation of mutations remains largely unknown. The introduced computational model demonstrates that three-dimensional (3D) crypt simulations are capable of improving the understanding of these interdependencies. We exemplify the potential of our model by simulation of a three-dimensional in silico model of mouse small intestinal crypts. Thereby, we focus on the intestinal SC niche being composed of undifferentiated functional SCs and Paneth cells (PC). In simulation series we demonstrate that both cell distribution and niche composition have strong impact on the clonal competition. We show that changes of the PC distribution can nearly abolish monoclonal conversion in the crypt. In addition we analyze the impact of Notch-signaling on clonal competition in the crypt. Intrinsic activation of Notchsignaling, i.e. autonomous Notch, in SCs suppresses specification of them into the PC lineage and eventually leads to PC depletion. Accordingly, these SCs can robustly maintain their state and thus gain a competitive advantage compared to cells that rely on wild type intercellular Notch activation. We find that under autonomous Notch monoclonal conversion in the crypts is accelerated. Moreover, quantifying the frequencies that Wnt-activating mutations become fixed in intestinal crypts, we observe that these frequencies are strongly increased in case of an autonomous Notch background compared to a wild type background. As a consequence early states of tumor development as metaplasia and adenoma can occur more frequently. Thus, we suggest that Notch plays a key role in clonal competition in the intestine. OP 16 Heterogeneity and cell fate control in mouse embryonic stem cells Herberg, M.1, Zerjatke, T.1, de Back, W.1, Kalkan, T. 2, Smith, A. 2, Roeder, I. 2, Glauche, I.1 TU Dresden, Institute for Medical Informatics and Biometry, Dresden, Germany Cambridge Stem Cell Institute, Cambridge, UK, Germany 1 2 Bistability is a characteristic feature of many molecular switches facilitating cell decision processes as they occur during differentiation and in response to external stimuli. Undifferentiated mouse embryonic stem cells (ESC) are a typical example, in which such a bistable situation is caught in an in vitro setting, thus contributing to a heterogeneous albeit dynamically stabilized cell culture. In particular, it is the expression of pluripotency factors like Nanog and Rex1 that obeys a bimodal distribution, which is reestablished after cell sorting. We studied heterogeneity and cell fate control in ESC using a variety of methods including cell culture experiments, flow cytometry, live-cell imaging, and quantitative image analysis. Complementary we developed a multiscale, spatial mathematical model of ESC growth, which allows comparing experimental results with our model predictions. The integrated, multiscale model serves as a framework to link population features, such as proliferation rates and spatial arrangements. Moreover, it reveals potential transcription factor related cellular and intercellular mechanisms behind the emergence of observed patterns that cannot be derived from experimental data directly. 42 OP 17 Multiscale Model of Liver Intoxication after APAP overdose. Boissier, N.1, Celliere, G.1, Ghallab, A. 2, Vignon-Clementel, I.1, Hengstler, J. 2, Drasdo, D.1 1 2 Sorbonne Universités, INRIA, UPMC Univ Paris 06, Lab. J.L. Lions, PARIS, France IfADo, Systemtoxikologie, DORTMUND, Germany Question: Paracetamol (acetaminophen, APAP) damage is the main cause of acute liver failure in many countries, including UK and the United States. Hepatocytes detoxify the blood from the drug. Above a certain threshold concentration, cells accumulate too high concentrations of NAPQI, a product of intracellular detoxification,and die. With increasing drug doses, an increasing necrotic lesion is generated. Liver zonation leads to different affinities and sensitivities of hepatocytesto APAP, depending on their spatial position along the portal-central lobule axis. The cells closest to the central vein of each lobule, the anatomical unit of liver, are the most sensitive to APAP and therefore the first to die. This impairs the ability of the liver to detoxify blood from ammonia with the threat of encephalopathy, and eventually death. We here address the question of how adverse effects of paracetamol can be predicted in a multiscale model that is calibrated with in vitro dose-response data using paracetamol as an example. Methods: We consider a spatial-temporal model of APAP intoxication in a liver lobule generated from the image analysis of mice liver using the software TiQuant[1], [2]. Blood flow is modeled as a resistive network taking into account the effect of red blood cells by using Pries’ law [3] for viscosity. Transport of APAP, modeled by a 1D transport PDE in the blood, is linked with its metabolism in the cells. The multi-scale model integrating the APAP detoxification pathway in each individual hepatocyte takes into account the spatial inhomogeneities within the liver and the impact of the temporal administration. The drug metabolic pathway of APAP has been calibrated by direct comparison with in vitro data. Results and Conclusions: Our approach allows us to couple straight forwardly the cells’ positions and the concentration of APAP that those cells locally experience taking into account cell-cell variability. Animal experiments are being done in order to validate simulation results for various doses. This work, besides modeling the acute liver intoxication after an overdose of paracetamol shows how in vitro experimental results and an in silico model of the liver can be combined to simulate the effect of a drug in vivo. [1] Hammad S., Hoehme S., Friebel A., et. al. “Protocols for staining of bile canalicular and sinusoidal networks of human, mouse and pig livers, three-dimensional reconstruction and quantification of tissue microarchitecture by image processing and analysis.” Archives of Toxicology 88 (5) 1161-1183 (2014) [2 ] Hoehme S., Brulport M., et al. “Prediction and validation of cell alignment along microvessels as order principle to restore tissue architecture in liver regeneration.” PNAS 107.23 10371-10376 (2010) [3] Pries A. R., Secomb T. W., Gaehtgens P., “Biophysical aspects of blood flow in the microvasculature.” Cardiovascular research. , 32(4), 654-667 (1996) OP 18 Think Adjoint - Methods facilitating parameter estimation for genome-scale mechanistic dynamic models Fröhlich, F.1,2, Shadrin, A. 3, Kessler, T. 3,4, Wierling, C. 3,4, Lange, B. 3,4, Theis, F.1,2, Lehrach, H.4,5, Hasenauer, J.1,2 Helmholtz Zentrum München, Institute of Computational Biology, Neuherberg, Germany Technische Universität München, Chair of Mathematical Modeling of Biological Systems, München, Germany 3 Alacris Theranostics, Berlin, Germany 4 Max Planck Institute for Molecular Genomics, Berlin, Germany 5 Dahlem Centre for Genome Research and Medical Systems Biology, Berlin, Germany 1 2 Main Objective: The progress of human diseases is influenced by a plethora of different molecular mechanisms. For most diseases it is difficult to identify key regulatory mechanisms as not all molecular species can be observed experimentally. Dynamic, mechanistic models can predict the dynamics of experimentally non-observable states, which allows for a more holistic understanding of the underlying mechanisms. However, for meaningful predictions the parameters of the dynamic model have to be estimated from experimental data. This process of estimating parameters from experimental data can be computationally challenging, especially for models describing interactions from pathway-scale to genome-scale where thousands of parameters have to be estimated. In this project we developed techniques that render parameter estimation for such large-scale dynamic models tractable. Methods: In many parameter estimation algorithms gradient-based optimization schemes are employed. Usually, gradients are computed using forward sensitivity equations or finite differences. These approaches scale poorly with model complexity and become prohibitively time consuming for genome-scale models. Instead, we propose the use of adjoint sensitivities analysis as a more efficient way to compute sensitivities for large-scale differential equation models. Results: We find that by using adjoint sensitivities we can reduce the computational burden of simulating genome-scale model and respective sensitivities by multiple orders of magnitude compared to finite difference and forward sensitivity based approaches. Furthermore, we estimated parameters for a genome-scale model from experimental data from multiple 43 Program / Abstracts thousand experimental conditions and showed that estimation time could be reduced to a small fraction of the previously required time. Conclusion: These results demonstrate that the developed techniques facilitate parameter estimation for pathway-scale to genome-scale models with thousands of state variables and thousands of parameters. This suggests the feasibility of predictions from dynamic mechanistic models trained on highly multiplexed data, such as transcriptomics or proteomics datasets, which were previously deemed to be intractable. OP 19 Development of an integrative model to improve anemia treatment in non-small cell lung carcinoma patients Rodriguez-Gonzalez, A.1,2,3, Schelker, M.4,5, Raue, A. 5,6, Steiert, B. 5, Böhm, M. E.1, Salopiata, F.1,3, Adlung, L.1, Stepath, M.1, Depner, S.1,2,3, Wagner, M.- C.1, Merkle, R.1,2, Kramer, B. A.1, Lattermann, S.1, Wäsch, M.1,3, Franke, A.7, Klipp, E.4, Wuchter, P. 8, Ho, A. D. 8, Lehmann, W. D.1, Jarsch, M.7, Schilling, M.1, Timmer, J. 5,9, Klingmüller, U.1,2,3 German Cancer Research Center (DKFZ), Systems Biology of Signal Transduction, Heidelberg, Germany Bioquant, Heidelberg University, Systems Biology of Signal Transduction, Heidelberg, Germany 3 German Center for Lung Research (DZL), Translational Lung Research Center (TLRC), Heidelberg, Germany 4 Institute of Biology, Humboldt-Universität zu Berlin, Theoretical Biophysics, Berlin, Germany 5 Institute of Physics, University of Freiburg, Freiburg, Germany 6 Merrimack, Discovery Division, Cambridge, United States 7 Roche Diagnostics GmbH, Roche Innovation Center , Pharma Research and Early Development (pRED), Penzberg, Germany 8 Heidelberg University, Department of Medicine V, Heidelberg, Germany 9 University of Freiburg, Centre for Biological Signalling Studies (BIOSS), Freiburg, Germany 1 2 Background: Cancer associated anemia is a common complication affecting 40% of all cancer patients during the course of the disease. It is particularly prevalent in lung carcinoma, with an incidence of 50-70% during disease development and reaching ≤90% in the advanced stages. Tumor inflammation and myelosuppressive chemotherapy are two of the major causes of anemia, reducing the therapeutic response and the quality of life in patients. Erythropoiesis Stimulating Agents (ESAs) have been widely used to reestablish normal levels of erythrocytes. However, up to 30-50% of cancer patients do not respond to the ESA treatment in cancer and several clinical trials reported higher mortality risk in ESA treatment. Furthermore, low levels of erythropoietin receptor (EpoR) protein expression has been found in tumor cell lines, rising concerns on EpoR activation in tumor context by ESA treatments. The low efficiency of anemia treatment and the safety concerns, indicate that new strategies are urgently required to improve the treatment options. Method: To transfer the dynamics of the interaction of the ligand erythropoietin (Epo) and the EpoR in the tumor and hematopoietic context, we combined mathematical modeling with quantitative data from pharmacokinetic and pharmacodynamic experiments of ESAs in human subjects, ESAs depletion in human erythroid progenitors and Non-Small Cell Lung Carcinoma (NSCLC) cell lines and the activation of signal transduction through the EpoR by mass spectrometry. The experimental data was used to establish an integrative mathematical model utilizing coupled ordinary differential equations (ODE) that links the cellular scale with the body scale. Results: The ODE model was able to describe the dynamic interaction of all ESAs at molecular, cellular and systemic level in the human body. Further, the ODE model facilitated to estimate the binding properties of all tested ESAs, to predict personalized optimal and safe dosage protocols for each ESA to activate the EpoR in the hematopoietic context but not in the tumor context. Conclusion: This model can describe the binding properties, dynamic interaction and the pharmacokineticpharmacodynamics of any EpoR ligand, predicting eficient and safe range of ESA concentrations in Lung cancer patients. OP 20 An expandable, multi-level, and multi-scale model for drug simulations of weight-loss and type 2 diabetes Nyman, E.1, Strålfors, P.1, Cedersund, G.1 1 Linköping University, Biomedical Engineering, Linköping, Sweden Question: Obesity and the metabolic syndrome are multi-level (intra-cellular to whole-body) and multi-scale (seconds to years) diseases, and to properly understand the action of associated drugs, a multi-level and multi-scale understanding is required. Such an understanding can only be obtained using systems pharmacology approaches, combining large amounts of data with mathematical modelling. However, all data cannot be considered simultaneously, since the complete modelling problem is too big. We therefore present an iterative approach, where sub-systems first are examined in isolation, and then integrated into a growing understanding of the whole. Methods: Models on all levels are described using ordinary differential equations and implemented in matlab. On the wholebody level, the sub-systems (e.g. the organs) are experimentally characterized in terms of their input-output profiles, which 44 allows the sub-systems to be modelled in isolation [1]. On the intracellular level, data include time-series and dose-responses for cellular metabolism and metabolic regulation, including mass spectrometry data describing the entire phosphoproteome [2-4]. Results: Through >10 iterations between experiments and modelling, we have unravelled the key mechanistic difference between normal healthy signaling, and insulin resistance in adipocytes: a feedback between the two proteins mTORC1 and IRS1 [3]. Apart from this feedback and a few already established protein abundances, the complete model explains both the normal and diabetic state using the same parameters (Fig 1)[3-4], and can describe all 30 000 phosphorylation sites measured in response to insulin stimulation. The model has been validated in several independent validation experiments, without parameter refitting, and can also translate the intracellular metabolic regulation to whole-body glucose homeostasis, describing both short-term (minutes and hours), and long-term changes (months and years) (Fig 2)[3-4]. Conclusions: We now have a multi-level model skeleton for several of the key changes involved in the development of type 2 diabetes. To this skeleton, new data and insights are continuously being added. The model and the modelling methodology are already used extensively by big drug development companies. [1] Nyman et al, J Biol Chem, 286(29):26028-41, 2011 [2] Humphrey et al, Cell Metab, 17(6):1009-20, 2013 [3] Brännmark et al, J Biol Chem, 288(14):9867-80, 2013 [4] Nyman et al, J Biol Chem, 289(48):33215-30, 2014 OP 21 Model-guided target identification for synergistic combination therapies in the DNA damage response pathway Raue, A.1, Alkan, O.1, Koshkaryev, A.1, Drummond, D.1, McDonagh, C.1, Schoeberl, B.1 1 Merrimack, Discovery, Cambridge, United States Cancer cells that are exposed to genotoxic stress activate signaling pathways that give rise to a variety of cellular responses such as DNA damage repair, cell cycle arrest and apoptosis. The relationships between signaling and these cellular responses are complex and not fully understood. In this work we present, for the first time, a computational model that quantitatively links the DNA damage signaling pathway response to cellular responses. In turn, these cellular responses are linked back to the origin of the DNA damage, creating a closed loop control system. We systematically investigate potentiating drug combinations in vitro and in vivo between DNA damage inducing chemotherapy and DNA damage signaling modulators and trained and validated our computational model with an extensive signaling data-set and time-lapse microscopy tracking the cell fate. The resulting trained computational model can predict the cellular responses based on alteration in DNA damage signaling pathway. We show how the model can be used to identify drug targets within the DNA damage repair pathway that are synergistic with different classes of DNA damaging agents. 45 Program / Abstracts OP 22 Identification of genetic determinants of immune cellular homeostasis using genetically diverse mouse strains Dubovik, T.1, Starosvetsky, E.1, Ziv-Kenet, A.1, Normand, R.1, Asbeh, N.1, Admon, Y.1, Shen-Orr, S.1 Technion-Israel Institute of Technology, Immunology, Haifa, Israel 1 Main objective: The immune system is one of the major life-sustaining systems in advanced organisms, employing dozens if not hundreds of distinct cell types each with their own function to orchestrate defense mechanisms. The immune system maintains cells in homeostasis, that is a relatively conserved stoichiometry which is altered during disease, and is returned to following. Baseline variation in cellular homeostasis exist between individuals which contributes to differences in immune response and is due both to environmental and complex genetic factors, but exhaustively mapping these to date has proven difficult. Here we set out to identify the genetic factors effecting individual level differences in baseline cellular immune homeostasis. Materials and methods: To address this question, we coupled a highly genetically diverse set of mouse models, the Collaborative Cross, which allow for high resolution genetic mapping; with mass-cytometry (CyTOF) a high dimensional single cell profiling technology which allows for high dimensional characterization of hundreds of thousands of cells in a sample. We profiled over 80 mice samples from 38 strains in bone marrow and spleen using CyTOF technology. We analyzed the data using high resolution clustering, followed by QTL analysis. Results: We identified a large number of genes that are significantly associated with observed variation in cell subset frequency across samples. Gene enrichment analysis revealed these are strongly enriched for cell death and survival genes which is supported by differences in proliferation rates in dividing cells between mice. Knowledge of which genetic variants of these genes a mouse carries allows prediction of immune cell subset frequency in an untested mouse. Conclusions: Cellular homeostasis is a complex trait whose maintenance of and fine tuning is performed through a delicate balance of genetic variants of cell production and turnover. OP 23 Assessing aneuploidy heterogeneity using single cell sequencing Taudt, A.1, Bakker B.1, Ferronika, P. 2, Belderbos, M. E.1,3, van den Bos, H.1, Porubsky, D.1, Spierings, D. C. J.1, Saber, A.4, de Jong, T. V.1, Halsema, N.1, Kazemier, H. G.1, Hoekstra-Wakker, K.1, Hiltermann, T. J. N. 5, Kok, K. 2, van der Wekken, A. J. 5, Timens, W.4, Bradley, A.6, E. de Bont, S. J. 3, Guryev, V.1, H. Groen, J. M. 5, Lansdorp, P. M.1, van den Berg, A.4, Foijer, F.1, Colomé Tatché, M.7,1 University Medical Centre Groningen, ERIBA, Groningen, Germany University Medical Centre Groningen, Department of Genetics, Groningen, Netherlands 3 University Medical Centre Groningen, Department of Paediatrics, Groningen, Netherlands 4 University Medical Centre Groningen, Department of Pathology & Medical Biology, Groningen, Netherlands 5 University Medical Centre Groningen, Department of Pulmonary Diseases, Groningen, Netherlands 6 Wellcome Trust Sanger Institute, Hinxton, Cambridge, Great Britain 7 Helmholtz Zentrum München, Institute of Computational Biology, Neuherberg, Germany 1 2 Whole chromosome duplications and deletions (aneuploidies) and copy number variations (CNVs) have been associated with many diseases, most prominently cancer. In cancer generally multiple aneuploidies and CNVs coexist in the same tumor, and this heterogeneity level may influence the metastatic potential of cancers and may determine the success of specific treatments. Quantifying the level of karyotype heterogeneity is therefore crucial for understanding the relationship between chromosomal instability and disease, and will help in the design of personalized treatments. Unfortunately, classic genomic profiling by comparative genomic hybridization (CGH) or whole-exome sequencing (WES) of bulk material can only identify global genomic alterations of a cell population. Single-cell whole genome Next Generation Sequencing (NGS) techniques offer a powerful way to study CNVs in a highthroughput fashion at a single cell level with high resolution. To analyze this single cell sequencing data we have developed AneuFinder, a method for the accurate and fully automated mapping of CNVs from single-cell NGS data which explicitly models the data structure of single cell NGS experiments by using a Hidden Markov Model where every distinct copynumber is assigned to a hidden state, and where read count distributions are modeled as negative binomial distributions. The resolution of this approach is only limited by sequencing coverage. We applied this method to map and quantify karyotype heterogeneity in mouse and human cancer samples. We used a mouse model of chromosome instability and show that the arising tumors exhibit high-grade karyotype heterogeneity, even though they tend to converge to particular and recurring chromosome combinations. These results indicate that selection forces for particular chromosome alterations can outcompete high missegregation rates. We also assessed CNVs in human primary small cell lung cancer tumors and corresponding liver metastasis, and found presence of rare cells in the primary tumor samples that showed CNV patterns identical to the metastases CNV pattern, suggesting that only specific cells or subclones of the primary tumor are responsible for the metastatic lesions. As chromosomal instabilities may drive tumour evolution, karyotype analysis using single-cell sequencing technology could very well become an essential tool for cancer treatment stratification in the near future. 46 POSTER PRESENTATIONS PP 1: Image-based Systems Biology PP 1-01 Label-free cell cycle analysis for high-throughput imaging flow cytometry Blasi, T.1, Hennig, H. 2, Summers, H. D. 3, Theis, F. J.1, Cerveira, J.4, Patterson, J. O.4, Davies, D.4, Filby, A. 5, Carpenter, A. E. 2, Rees, P. 2,3 ICB Helmholtz Zentrum München, Neuherberg, Germany Broad Institute, Cambridge, Great Britain 3 Swansea University, Swansea, Great Britain 4 The Francis Crick Institute, London, Great Britain 5 Newcastle Upon Tyne University, Newcastle Upon Tyne, Great Britain 1 2 Imaging flow cytometry combines the high-throughput capabilities of conventional flow cytometry with single-cell imaging. Here we demonstrate label-free prediction of DNA content and quantification of the mitotic cell cycle phases by applying supervised machine learning to morphological features extracted from brightfield and the typically ignored darkfield images of cells from an imaging flow cytometer. This method facilitates non-destructive monitoring of cells avoiding potentially confounding effects of fluorescent stains while maximizing available fluorescence channels. The method is effective in cell cycle analysis for mammalian cells, both fixed and live, and accurately assesses the impact of a cell cycle mitotic phase blocking agent. As the same method is effective in predicting the DNA content of fission yeast, it is likely to have a broad application to other cell types. PP 1-02 Cells commit to the cell cycle by rapid and irreversible inactivation of APCCdh1 Cappell, S.1, Chung, M.1, Jaimovich, A.1, Spencer, S.1, Meyer, T.1 1 Stanford University, Chemical and Systems Biology, Stanford, United States The current model of the mammalian cell cycle distinguishes a single commitment point, when cells can no longer exit to quiescence, from multiple subsequent checkpoints, when committed cells pause on their path to complete the cell cycle. A commitment point often consists of a bistable mechanism that confers strong hysteresis, meaning a different sensitivity to inputs before and after the switch. We used live-cell time-lapse imaging of cell cycle biosensors and single-cell image analysis to identify the mechanism underlying cell cycle commitment. Here we demonstrate that inactivation of the anaphase promoting complex/cyclosome (APC-Cdh1) constitutes a bistable switch in late G1-phase that is initiated by cyclin E/Cdk2 and made irreversible by Emi1 with strong hysteresis in respect to stress sensitivity and Cdk2 inactivation. The switch constitutes a bona fide commitment point, as exposure to stress until but not after APC-Cdh1 inactivation reverted cells to a mitogen-sensitive quiescent state allowing for later re-entry into the cell cycle. The switch is distinct from activation of the Rb-E2F pathway earlier in G1, as stress-sensitivity persisted for many hours after Rb hyperphosphorylation and E2F activation; and is also distinct from the checkpoint at the G1/S boundary, which pauses cells at the onset of S-phase only after APC-Cdh1 inactivation. Thus, APC-Cdh1 inactivation constitutes a bistable switch with the necessary characteristics to be the commitment point for cell cycle entry. PP 1-03 Image-based systems medicine: from mechanistic models to decision-support systems in the clinic Sten, S.1,2, Karlsson, M.1,2, Forsgren, M. 2, Lundengård, K.1,2, Lundberg, P. 2, Engström, M. 2, Cedersund, G.1 1 2 Linköping University, Biomedical Engineering, Linköping, Sweden Linköping University, Medicine and Health, Linköping, Sweden Question: Clinical diagnosis is today often based on images from magnetic resonance imaging (MRI) cameras. Such images can characterize e.g. fat infiltration in the liver and areas of activity in the brain (fMRI). However, today, such characterizations are done using simple phenomenological approaches, such as quantifications of amplitudes, normalizations, and correlations with basis functions. In this talk, we will demonstrate that data analysis is improved by usage of mechanistic models. Methods: Models were formulated as ordinary differential equations in matlab; parameters were inferred from data within physiological ranges using nonlinear mixed-effects modelling; and predictions were obtained with uncertainty by considering all feasible parameters. Data are obtained from healthy subjects and patients: liver patients have suspected nonalcoholic fatty liver disease, brain patients have stroke. Time-series measurements for the liver measure the uptake of a liverspecific contrast agent, and brain measurements detect blood oxygenation in response to visual stimuli and tasks. Results: The models can describe training data and independent validation data without parameter refitting [1-2] (Fig 47 Program / Abstracts 1A,B). Key parameters estimated with low uncertainty provide new and clinically valuable information: liver transport rate parameters are able to predict the degree of fibrosis seen in liver biopsies better than any other previously suggested method for analysis of MRI data, and model-based brain activity and blood-flow parameters are more accurate and reliable estimates of these entities compared to corresponding estimates obtained using clinically used software (Fig 1C). Conclusions: We have demonstrated that the mechanistic process understanding that goes into our mechanistic models allow us to provide a superior image analysis in terms of personalized diagnosis. This demonstrates that some mechanistic models now are ready for implementation in end-usage decision-support systems. We end by a discussion of the big changes in our healthcare systems that such implementations will imply, and how we combine competences of widely different characters to soundly approach these changes (Fig 2). This discussion will also illustrate that our mechanistic models open the door to multi-level decision-support models and a combination of radiology and omics, sometimes referred to as radiomics (Fig 3). [1] Forsgren et al, PLoS One, 9(4):e95700 [2] Lundengård et al, submitted PP 1-04 Investigation of Bile canaliculi formation and biliary transport in 3D in vitro liver cultures Damm, G.1, Deharde, D.1, Schneider, C. 2, Hiller, T. 2, Kegel, V.1, Lübberstedt, M. 2, Freyer, N. 2, Seehofer, D.1, Zeilinger, K. 2 Charité - Universitätsmedizin Berlin, Klinik für Allgemein-, Visceral- und Transplantationschirurgie, Berlin, Germany Charité - Universitätsmedizin Berlin, Bioreactor Group, Berlin-Brandenburg Center for Regenerative Therapies (BCRT), Berlin, Germany 1 2 Background: Primary human hepatocytes (PHH) are still considered as gold standard for investigation of in vitro metabolism and hepatotoxicity. It has been shown that the three- dimensional (3D) cultivation of PHH in a sandwich configuration between two layers of extracellular matrix (ECM) enables the hepatocytes to adhere three-dimensionally leading to formation of in vivo like cell-cell contacts and cell-matrix interactions. In the present study the influence of the matrix composition on hepatocyte morphology and functionality was analysed applying a systems biological approach. Methods: Human hepatocytes were isolated from human liver resectates using a two-step collagenase perfusion technique. Freshly isolated PHH were cultured for 6 days between two ECM layers made of collagen and/or Matrigel in four different combinations. The cultures were investigated by phase contrast microscopy and immunofluorescence analysis with respect to cell-cell connections, repolarization as well as bile canaliculi formation. Finally, the bile canalicular transport was analyzed by live cell imaging to monitor the secretion and accumulation of the fluorescent substance CDF in bile canaliculi. Large image data sets were analyzed using biostatistical and bioinformatic methods. Results: Investigated sandwich cultured PHH showed ECM dependant differences in morphology and cellular arrangement. Cultures with an underlay of collagen seem to represent best the in vivo tissue architecture in terms of formation of 48 trabecular cell arrangement. Cultures overlaid with collagen were characterized by the formation of abundant bile canaliculi, while the bile canaliculi network in hepatocytes cultured on a layer of Matrigel and overlaid with collagen showed the most branched and stable canalicular network. All cultures showed a time-dependent leakage of CDF from the bile canaliculi into the culture supernatant with variations of CDF half-life in dependence of the used matrix combination. Conclusion: In conclusion, the results of this study show that the choice of ECM has an impact on the morphology, cell assembly and bile canaliculi formation in PHH sandwich cultures. The morphology and the multicellular arrangement were essentially influenced by the underlaying matrix, while bile excretion and leakage of sandwich cultured hepatocytes were mainly influenced by the overlay matrix. Leaking and damaged bile canaliculi could be a limitation of sandwich culture models in long-term excretion studies. PP 1-05 Imaging the dynamics of the proteome in response to viral infection Drayman, N.1, Alon, U.1 Weizmann Institute of Science, Cell Biology, Rehovot, Israel 1 Individual cells in a population can show different responses to stress stimuli. We explored how viral infection of human cells is affected by cell-to-cell variations in protein level and localization over time. To address this, we use a unique assay developed in our lab to follow ~500 different proteins, fluorescently tagged at their endogenous chromosomal locus. We determine the proteomic response of the cells to Herpes Simplex Virus 1 infection, a well-studied viral system and a major human pathogen. We follow the dynamics of proteins localization and level in each individual cell over time by timelapse microscopy, under conditions where some of the cells become successfully infected while others are able to resist the infection. This allows us to identify proteins whose differential dynamics at early times correlates with eventual infection success or failure. This study provides the first global view on how initial cell-to-cell variability correlates with the outcome of infection, allowing us to move from a “virus-centric” view, where each cell is equally likely to become infected, to a more complex “virus-host” view, where the probability of infection also depends on the cell state. PP 1-06 Functional intravital imaging of acetaminophen induced liver injury and regeneration Ghallab, A.1, Reif, R.1, Hassan, R.1, Seddek, A. 2, Hengstler, J. G.1 1 2 IFADo, Dortmund, Germany Faculty of Veterinary Medicine, South Valley University, Qena, Egypt, Egypt Background and objective: Acetaminophen (APAP) overdose is one of the most frequent causes of acute liver failure worldwide. Although the signaling pathways involved in cell death are extensively studied, the sequence of events during the destruction and the regeneration processes are not completely understood. Here intravital imaging using two-photon microscopy can play a role. It allows imaging of biological processes in a real time. The purpose of this study is to image in vivo the cell death and regeneration after APAP challenge. Materials and methods: Acute liver damage was induced by intraperitoneal administration of APAP (300 mg/kg). Intravital imaging was done using two-photon microscopy after administration of the following vital dyes: Hoechest, a DNA marker; Rhodamine (TMRE), a marker of mitochondrial membrane potential; Cholyl-Lysyl-Fluorescein (CLF), a fluorescein labeled bile salt) and propidium iodide, a marker of cell death. Immune cells infiltration was visualized using cell type specific reporter mice. Results: The earliest event, approximately 20 min after APAP injection, was loss of mitochondrial activity of pericentral hepatocytes as evidenced by TMRE loss. The compromised mitochondrial function was followed immediately by ballooning of the bile canaliculi. Approximately 1h later the widened bile canaliculi invaginated into the adjacent hepatocytes. 2h after APAP injection, hepatocytes with bile canalicular protrusions lost their ability to secrete bile salts, leading to irreversible cell death as evidenced by propidium iodide uptake. Interestingly, this bile salts mediated cell death occurs only at the outer pericentral hepatocytes. In contrast, hepatocytes which are in close contact to the central vein showed nuclear fragmentation. This destruction process is followed by a well-orchestrated regeneration process characterized by stellate cells activation and immune cells infiltration. Our imaging revealed that macrophages infiltrate the dead cell area only after day two following APAP injection. Although work for a formal proof is still in progress, it seems as if the infiltrating macrophages interact with and finally phagocytose stellate cells in the dead cell area. This was supported by the prolonged activation of stellate cells when we depleted macrophages by clodronate administration Conclusion: The direct observation of cellular and sub-cellular events in the living liver allows insights into the sequence of pathophysiological events which are difficult to obtain by conventional methods. 49 Program / Abstracts PP 1-07 A Systems Survey of Progressive Host Cell Reorganization During Rotavirus Infection Green, V.1, Pelkmans, L.1 University of Zurich, Institute of Molecular Life Sciences, Zurich, Switzerland 1 Question: Pathogen invasion is often accompanied by widespread alterations in cellular physiology, which reflects the hijacking of host factors and processes for entry and replication. Although genetic perturbation screens have revealed the complexity of host factors involved for numerous pathogens, it has remained challenging to disentangle this complexity along the progression of host cell reorganization during the infection process. Methods: We here address this by combining the first high confidence, image-based, genome-scale RNAi screening of rotavirus infection in human intestinal cells with an innovative approach to infer a continuous trajectory of virus infection progression from fixed cell populations. Results: All five large-scale RNAi screens performed were reproducible and showed a high degree of consistency. Through a probabilistic method for the aggregation of multiple datasets, we obtained a gene score based on their likelihood to confer an on-target infection phenotype upon RNAi, within which known rotavirus host factors were retrieved, validating our approach. By developing a novel, rank-based analysis for functional annotation enrichment, we provide an unbiased, systems-wide view of the cellular processes important for rotavirus replication. The result is a wealth of newly identified host factors in rotavirus infection, providing a rich resource for the field. Combined with our multivariate, single-cell infection progression trajectory, we uncover a complex, yet ordered host cell reorganisation program during rotavirus infection that provides a replication-permissive cellular environment for the virus. This includes an alternative mechanism for host protein synthesis shut-off during early stages of infection involving mTORC1 inhibition, consumption of lipid stores, rearrangement of mitochondria, and ER shape remodelling throughout infection, modulated by the ER shaping protein REEP2. Finally, by integrating the large-scale gene perturbation dataset with single-cell trajectories, we propose a model for host factormediated cellular reorganization during rotavirus infection, many elements of which can be linked to calcium-activated AMPK signalling. Conclusions: Our work provides a powerful approach to order the complexity of host cellular requirements along a trajectory of cellular reorganisation during pathogen invasion, an approach that would enhance the analysis of perturbation screens from any field. 50 PP 1-08 A Qualitative Method of Identifying Stiff Regions in Fibrous Proteins Horan, S.1, Kurup, A. 2, Botvinick, E. 2 University of California, Irvine, Mathematics/Center for Complex Biological Systems, Irvine, United States University of California, Irvine, Biomedical Engineering, Irvine, United States 1 2 Fibrous proteins such as collagen and fibrin are ubiquitous in both cells and the extra cellular matrix surrounding them in complex organisms. Examining them from an engineering perspective gives rise to questions regarding their mechanical properties, particularly their behavior under stress. We present a method for using MATLab to identify regions of relative stiffness and flexibility in such proteins by analyzing images from confocal microscopy. This method uses five major steps: First, a series of images of these fibers is taken over a short time. Second, areas containing fibers are identified and skeletonized to locate the likely position of the actual fiber. Third, a mask using this skeletonization is applied to each image, leaving only pixel/voxel data where we believe the fiber to be located. Fourth, each pixel/voxel remaining is considered as a stack over time and cross correlated with other itself pixel/voxel stacks. Finally, regions of close correlation to the original stack are identified, and this region size is considered to be an indication of stiffness. When testing this method on images of fibrin, we found regions of relative stiffness only where fibers were straight, but the converse did not hold. We consider this to be in support of our method, as one would expect that stiff regions are straight, but straight regions need not necessarily be stiff. We also found that regions where the fibers are seen to bend are identified as less stiff, which is again to be expected. Interestingly, while we started with image stacks of 100 frames for analysis, we found results to be qualitatively similar when using only much smaller Our image analysis current works in two dimensions, but we hope to have three ready and tested by the time of the conference. In the future, we plan to extend this work by examining these proteins under various stresses and compare these results to materials with known bending modulus to make more quantitative judgments. PP 1-10 Image-based modeling of organogenesis Lang, C.1, Michos, O.1, Iber, D.1 1 ETH Zurich, D-BSSE, Basel, Switzerland To achieve a high surface-to-volume ratio, branched organs such as the lung and the kidney undergo a complex, yet highly stereotyped morphogenesis process. It is an open question how the branching program can be both stereotypic, yet capable to adjust to the available space. The molecular players underlying bud specification and outgrowth have been extensively studied. Thus, FGF10 and GDNF signalling are both necessary and sufficient to guide the outgrowth of new branches in the lung and kidney respectively. Several mechanisms have been proposed to explain how signalling can self-organize into spots to define new branch points. We have combined 3D microscopy and image-based spatiotemporal modeling to test the validity of the proposed models. We find that of all proposed mechanisms only a receptor-ligand based Turing-type model can reproduce the measured embryonic growth fields and guide the outgrowth of new branches to recapitulate the embryonic 3D branching process (Menshykau et al, 2014, Development; Georgieva et al, in preparation). The ligand-receptor based Turing mechanism recapitulates all reported qualitative lung and kidney branching mutant phenotypes (Celliere et al, 2012). Additionally, we have now used 2D 51 Program / Abstracts lung and kidney cultures to quantitatively test the ability of the Turing mechanism to recapitulate the branching behaviour in mutants and in the presence of inhibitors (Menshykau et al, in preparation; Kull et al, in preparation). The ligand-receptor based Turing mechanism not only recapitulates all data, but also offers an explanation of how stereotyped patterning can be achieved in spite of molecular noise and how a space-filling tree can be generated (Menshykau et al, 2014, Development; Georgieva et al, in preparation). In conclusion, image-based spatiotemporal modeling allows us to identify, test, and refine mechanisms to explain the control of the complex branching program during mouse lung development. PP 1-11 Clustering-based classification of autophagy phenotypes in single cell images as a novel readout of autophagic activity Marin Zapata, P. A.1, Hamacher-Brady, A.1 1 DKFZ, Lysosomal Systems Biology, Heidelberg, Germany Autophagy is a cellular degradative pathway in which cytosolic components are engulfed in double membranes vesicles (autophagosomes) and subsequently delivered to the lysosomes. Quantification of autophagy in single cells is prone to misinterpretation, since increased autophagosome number could be indicative of increased autophagosome synthesis (autophagy activation) as well as blocked degradation (autophagy inhibition). Thus, the autophagy field is still in need for unbiased single-cell readouts. Imaging cytometry provides a platform for high throughput, single-cell image acquisition and analysis, which is potentially useful for autophagy quantification. However, the statistical power and high-content multiparametric nature of this technique have not been exploited within the context of autophagy. In this work, we developed a high content analysis for classification of autophagy phenotypes in imaging cytometry data based on agglomerative clustering of texture features. Our approach was able to automatically identify subpopulations of low, high and blocked autophagic flux in multiple cell lines. Furthermore, we propose convenient data representations of subpopulation abundances (denoted as fingerprints) to establish a subpopulation-based characterization of autophagy response. Besides quantifying cell heterogeneity, fingerprints provided unique identifiers for fully feed, nutrient deprived and lysosome inhibited conditions, and correctly reported the magnitude and direction of autophagic flux in terms of number of cells. Thus, we propose our analysis as an unbiased autophagy readout of potential use for the characterization of new compounds. PP 1-12 MALDI IMAGING OF A PROSTATE CANCER BIOPSY - A PILOT STUDY Rise, K.1, Tessem, M.- B.1, Nordgård, A. 2, Borgos, S. E. 2, Bathen, T. F.1, Andersen, M. K.1, Drabløs, F.1, Rye, M.1, Bertilsson, H.1 1 2 Norwegian University of Science and Technology, Trondheim, Norway SINTEF, Trondheim, Norway Question: Prostate cancer is the most common cancer in men and the number of diagnosed patients increases every year, both as a result of men living longer and the increased usage of Prostate Specific Antigen (PSA) as biomarker. Since PSA is an uncertain marker, new methods for stratification at tissue level is in demand to avoid overtreatment of indolent disease. Matrix Assisted Laser Desorption Ionization (MALDI) Imaging is a novel tool for finding new biomarkers with better sensitivity and specificity for prostate cancer. By using a suitable matrix, analytes are extracted and a mass spectrum is generated for each measuring point in a grid system. This creates a high resolution spatial distribution of analytes across the tissue sample, where the confounding effects on the signal from artefacts and tissue composition can be easily assessed. MALDI Imaging can be important in three areas of early detection: biomarker identification and validation, improved interpretation of clinical images obtained from state-of-the-art magnetic resonance imaging, and improved technology with MALDI Imaging linking in vivo and ex vivo measurements of metabolism. Methods: Biopsies were frozen prior to embedding in a gelatin-cellulose mixture, and sliced using a cryostat with optimized conditions. Slices were placed on indium tin oxide (ITO) slides, and MALDI matrix 2,5-Dihydroxybenzoic acid (DHB) was added. Mass calibration was performed using matrix spots outside the sample area, and Bruker Solarix FT ICR was applied to obtain mass spectrometric images. Mass spectra were collected in positive mode in mass range 50 - 2000. Results: The average spectrum shows that most m/z values were observed around 650 - 950. The images showed a variety of masses, with colour intensity related to ion abundance, and it clearly shows how the masses are unevenly distributed in the samples. Most compounds have not yet been identified, although there is a possible match for spermine at m/z 203. Conclusion: The study shows a potential of using MALDI Imaging in the search for new and improved biomarkers and identification of metabolism in heterogene tissue of cancerous prostates. Further studies will aim to optimize parameters, including slice thickness, choice of matrix, pre-processing of slices and optimisation of mass spectrometric method. Further analysis will be performed to identify compounds in the samples. 52 PP 1-13 Complete evaluation of dynamics in an unlabelled live cell from a z-stack of superresolved bright-field microscopic images Rychtarikova, R.1, Nahlik, T.1, Stys, D.1 University of South Bohemia, Inst. of Complex Systems, Nove Hrady, Czech Republic 1 Question: Cellular dynamics determines key processes in development of organism, tissue healing and cancer invasion. We proposed an algorithm of processing of z-stacks of images of unlabelled live cells obtained from wide-field bright-field optical transmission microscopy to achieve the super-resolved volumes of intracellular objects in the order of 100 nm3, including the classification of pixels according to an intracellular dynamics and spectral properties [1,2]. Methods: This super-resolved method is based (1) on non-interpolating de-mosaicing of a 12-bit z-stack of raw images obtained from camera chip equipped by a Bayer mask, which preserves as much information in the image as possible to obtain RGB images. (2) on searching for pixels of unchanged (or near) intensities between two consecutive images, after either the simple subtraction of two consecutive images or the calculation of a point divergence gain: where k is the Renyi coefficient, j is the number of phenomena in the discrete distribution, pi and p k/l are the probabilities of occurrences of the i-th phenomenon in the original distribution and in the distribution after supplying one element of the phenomenon i = k by one element of the phenomenon i = l. The latter step of image processing is responsible for evaluation of the intracellular dynamics. If PDG = 0 (the same intensities at the same position in two consecutive images), large intensity homogenous non-moving objects are mainly segmented. At minimal and maximal values of PDG (a pixel of the rarest intensity is replaced by a pixel of the most frequent intensity and vice versa), we track a large moving organelles. The other values of PDG correspond either to the other intracellular objects or to other intensities in the course of the sum of point spread functions of the live cell. (3) the selection and evaluation of each colour channel of the image. In the blue channel, we observe mainly autofluorescence. In the green channel, the light diffraction is also projected. In the red channel, the light absorption in near infrared region is further observed. Results and Conclusion: The method was tested and verified on a few cell lines and pluripotent stem cells and shows also a potential to describe dynamics in primary cell cultures. The patent application on the algorithm and related microscope is also pending [3]. References: [1] R. Rychtáriková et al., in ISCS 2014: Interdisciplinary Symposium on Complex Systems (Emergence, Complexity and Computation 14), edited by A. Sanayei, O.E. Rössler, I. Zelinka (Switzerland: Springer), 2014, pp. 261-267. [2] Štys D., Rychtáriková R., Náhlík T., Urban J., Císař P.: Method of Microscopic Images Acquisition and Processing and the Particular Device, Patent Application No. PV 2015-434. PP 1-14 3D reconstruction and quantitative analysis of liver human samples Segovia-Miranda, F.1, Belicova, L.1, Morales-Navarrete, H.1, Seifert, S.1, Hampe, J. 2, Kalaidzidis, Y.1, Zerial, M.1 1 Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany 2 Medical Department 1, TUD, Dresden, Germany The liver has a complex 3D structure that can be hardly extracted from the classical 2D histological approaches. 3D analysis of the liver tissue should be taken into account for a better understanding of liver physiology and physiopathology. The aims of this study are 1) to establish a staining protocol to perform high-resolution confocal imaging on thick human liver slices and 2) to quantitatively describe the 3D architecture of human liver tissue. Human liver sections (100 um thick) were stained for cell borders (LDL receptor), bile canaliculi (CD13), sinusoids (goat anti-human A488), and nucleus (DAPI), optically cleared and imaged using multiphoton microscopy. The Motion Tracking software was used to segment and quantitatively describe 1) the bile canaliculi and sinusoidal networks and 2) the hepatocyte morphology. Our data showed that the set of antibodies and dyes chosen allows the 3D reconstruction of several components of liver architecture, including central vein, 53 Program / Abstracts portal vein, bile canaliculi, sinusoids, nuclei and hepatocytes. We provided a quantitative description of the bile canaliculi and sinusoidal networks (total length, connectivity, volume fraction, diameter, mean branch length and density). We also extracted precise quantitative morphometric parameters of liver tissue focusing on hepatocyte cells (e.g. hepatocyte volume, surface area, apical, lateral and basal surface, % mono-nuclear and bi-nuclear hepatocytes, number of neighbors). We believe the detailed structural information provided by the 3D analysis of liver architecture will reveal new aspects of organ physiology. This new tool will be used during the LiSyM funding period to image human samples of patients diagnosed with non-alcoholic fatty liver disease (NAFLD) to describe the morphological changes occurring in the tissue during disease progression. This will allow exploring new aspects of liver disease that have not been described before due to the technological challenge, thus contributing to the understanding of NAFLD liver physiopathology. PP 1-15 QUANTITATIVE ANALYSIS OF TUMOR CELL LOAD BASED ON DIFFUSION-WEIGHTED MRI AND HISTOLOGY DATA Yin, Y.1,2, Sedlaczek, O. 3,4,5, Müller, B.6, Lotz, J.7, Warth, A. 3,6, González-Vallinas, M.6, Lahrmann, B.6,8, Grabe, N.6,8, Kauczor, H.- U. 3,4,5, Olesch, J.7, Breuhahn, K.6, Vignon-Clementel, I. E.1,2, Drasdo, D.1,2,9 INRIA de Paris, Paris, France Sorbonne Universités UPMC Univ. Paris 6, Laboratoire Jacques-Louis Lions, Paris, France 3 Translational Lung Research Center Heidelberg (TLRC), member of the German Centre for Lung Research (DZL), Heidelberg, Germany 4 Department of Diagnostic & Interventional Radiology, University Hospital of Heidelberg, Heidelberg, Germany 5 Department of Diagnostic and Interventional Radiology with Nuclear Medicine, Thoraxklinik at University of Heidelberg, Heidelberg, Germany 6 Institute of Pathology, University Hospital Heidelberg, Heidelberg, Germany 1 2 Fraunhofer Mevis, Lübeck, Germany Department of Medical Oncology, NCT, University Hospital Heidelberg, Heidelberg, Germany 9 IZBI, University of Leipzig, Leipzig, France 7 8 Tumor cellularity is an important tissue microstructural feature, which is useful for cancer diagnosis and cell number related treatment such as targeted cancer therapies. Diffusion-weighted MRI (DWI) that indicates the rates of water diffusion in tissue is used in medical imaging as it is thought to provide an indirect measure of tumor cellularity. Histopathological examination of tissues reveals the tissue microstructure, but is usually available as a small sample or after resection. How to quantitatively estimate the total tumor cell load is still a challenging work. In this study, we proposed a pipeline for tumor cell load estimation based on image processing techniques. The main pipeline steps include: 1) adaptive cell nuclei segmentation; 2) model-based 3D cell density estimation; 3) correlation identification between the DWI diffusion coefficient (D value) and the underlying tumor cellularity; 4) cell number calculation based on tumor D values and the identified correlation between diffusion and cellularity. The proposed cell segmentation algorithm performs detection with a very high accuracy (0.95) on selected tumor tissue samples. For a specific patient with NSCLC, the estimated total cell number of the tumor (6.5 cm in diameter) is 3.64×1010. The proposed method translates tumor DWI signals into cell density and quantifies the whole tumor cell load. The developed method can be applied to other cancer types. PP 1-16 Investigating Microenvironment-to-cell Signaling in 3D Spheroids through Imaging Mass Cytometry Zanotelli, V.1, Georgi, F.1, Schapiro, D.1, Andriasyan, V.1, Yakimovich, A.1, Catena, R.1, Jackson, H.1, Bodenmiller, B.1 University of Zurich, Institute of Molecular Life Science, Zurich, Switzerland 1 Question: In vitro studies using monolayer 2D cell cultures inherently fail to model a physiological 3D setting, including 3D cell-to-cell contacts as well as nutrient and oxygen gradients. 3D tissue models such as spheroid cell cultures overcome this drawback and resemble the in vivo situation more faithfully. Consequently these tissue models are currently broadly adapted in biomedical and pharmaceutical research. Every cell senses its local 3D environment and adapts its phenotype accordingly. This process is a major driver for the regulated phenotypic heterogeneity in tissue development and homeostasis. When deregulated it can drive diseases such as cancer. However, how the interactions of cells with their environment collectively shape cellular phenotypes in tumors and contribute to tumor heterogeneity is largely unknown. In order to tackle this question we set out to develop a high throughput setup to quantify the influence of microenvironmental influences and cell-to-cell communication on the phenotypic heterogeneity in an in vitro 3D breast cancer cell culture system. Methods: Imaging mass cytometry (IMC) allows the simultaneous quantification of more than 40 phenotypic and functional markers at subcellular resolution in slices of 3D tissues. This makes this technology suitable to study the spatial relationships of complex phenotypes and how both, relationships and phenotypes, are perturbed by small molecule inhibitors. To enable screening approaches using 3D tissue models and imaging mass cytometry, we are developing a workflow based on metal label based barcoding that allows pooling of spheroids treated with a multitude of conditions before the embedding and 54 cutting steps for subsequent IMC. In the future the presented pipeline will allow for the concurrent embedding and cutting of a 384 well sphere culture plate with up to 384 different conditions, making a high throughput screen with 3D spheroids coupled with an IMC readout feasible. Results: We use this workflow to study the origins of phenotypic heterogeneity of 3D cultured breast cancer cell lines. Preliminary data assesses the feasibility and challenges a barcoding based approach to combine 3D spheroid cell culture technology efficiently with IMC. Further, based on data of unperturbed breast cancer spheroids we show how IMC can capture phenotypic heterogeneity and coordination in the microenvironment. We also explore how such data can be integrated using mathematical modeling to gain quantitative insights. Conclusions: We present the development of a broadly applicable, scalable screening approach efficiently combining high throughput 3D microtissue culture with IMC. The approach will also be applicable to more complex cell culture settings, such as 3D co-cultures and organoids and to advanced readouts such as 3D IMC. Combined with an inhibitor screen the technology will be a solid basis to investigate how phenotypes are spatially coordinated in 3D tumor models. PP 1-17 Analysing the Impact of Errors in Single Cell Tracking Experiments Zerjatke, T.1, Roeder, I.1, Glauche, I.1 TU Dresden, Institute for Medical Informatics and Biometry, Dresden, Germany 1 Time-lapse video microscopy is an increasingly popular method to study the temporal and spatial behaviour of single cells. It can be used for a broad range of applications, e.g. the analysis of cell motility and migration, proliferation properties, clonal composition, or the reconstruction of the complete divisional history of cells, represented as cellular genealogies. A large number of automated methods has been developed for segmenting and tracking single cells. Although these methods are increasingly sophisticated to cope with a broad spectrum of situations they inevitably produce errors in the reconstruction of cellular tracks. The number of errors can be reduced by using post- processing tools for the manual correction of automatically created tracks. However, ambiguous situations can occur that lead to different subjective decisions of individual raters in the assignment of cellular objects. The number of these ambiguous situations and hence the number of differences in the reconstructed cellular tracks depends on cell type specific properties like migration speed or proliferation rate, as well as on specific properties of the experimental setting like the spatial and temporal resolution of the image sequence or the density of seeded cells. Here we study this inter-rater variability exemplarily for in vitro cultures of haematopoietic stem and progenitor cells and analyse its impact on the reconstruction of cellular genealogies and statistical measures of e.g. migration properties. Furthermore, we use computer simulations of in vitro cell cultures that allow to comprehensively analyse a broad range of cell type specific and experimental properties. Specifically, we aim to quantify maximum error rates that are admissible to reliably measure a particular statistical outcome. These maximum admissible error rates can then be accounted for in the design of the experimental set-up and the choice of the cell tracking procedure. PP 2: Single-Cell Systems Biology PP 2-01 destiny - diffusion maps for large-scale single-cell data in R Angerer, P.1, Haghverdi, L.1, Büttner, M.1, Theis, F.1, Marr, C.1, Büttner, F. 2 1 2 Institite of Computational Biology, Neuherberg, Germany European Bioinformatics Institute, European Molecular Biology Laboratory, Cambridge, Great Britain Diffusion Maps (DMs) are a spectral method for non-linear dimension reduction. destiny is an efficient R implementation of the diffusion map algorithm, adapted for the visualization of single cell expression data. Those adaptions include a singlecell specific noise model allowing for missing and censored values, an efficient nearest-neighbour approximation that allows for the processing of hundreds of thousands of cells, and a functionality for projecting new data on existing diffusion maps. We show the application of destiny to 4 datasets of varying size and complexity to show both the capabilities of the implementation and the usefulness of the method and its adaptions. 55 Program / Abstracts PP 2-02 Sensitive Detection of Rare Disease-Associated Cell Subsets via Representation Learning Arvaniti, E.1, Claassen, M.1 1 ETH Zurich, Biology, Zurich, Switzerland Rare cell populations play a pivotal role in the initiation and progression of diseases like cancer. While increasingly multiparametric single-cell technologies are opening a new perspective to study well-characterized low frequency cell subsets, the systematic de novo identification of such disease-associated cell subsets remains an unsolved challenge. This work describes CellCnn, a supervised representation learning approach to infer and detect, possibly rare, diseaseassociated cell types from single cell measurements. Specifically, CellCnn learns an optimal representation of a set of single cell abundance profiles with respect to an associated phenotype by means of a convolutional neural network. The learned representation lends itself to identify molecular profiles of phenotype-associated cell subsets. We demonstrate the ability of CellCnn to correctly identify subsets of peripheral mononuclear blood cells with differential signaling response to various paracrine agents, and subsets associated with continuous valued clinical parameters such as disease onset in HIV infected individuals. We specifically assess the CellCnn capability to detect extremely rare leukemic blast cell populations in acute lymphoblastic leukemia and acute myeloid leukemia. While conventional cell sample classification approaches fail to identify the leukemia-associated cell subsets for any of the considered cell frequencies, CellCnn is able to faithfully achieve this task for cell subset frequencies as low as 0.005 %. CellCnn constitutes a dedicated supervised machine learning approach to identify rare phenotype-associated cell subsets from single cell measurements. With the dramatic increase in single cell studies of diverse cancers and other diseases, we expect CellCnn to enable de novo, disease- and possibly also patient-specific discovery of disease-associated cell populations and thereby contribute to the elucidation of disease mechanisms mediated by rare cell events and the design of cell subset specific treatment interventions. PP 2-03 Information theoretic analysis of interleukin-6-induced signalling by multi-colour flow cytometry Billing, U.1, Rummel, H.1, Schaper, F.1, Waldherr, S. 2, Dittrich, A.1 Otto-von-Guericke University, Department of Systems Biology, Magdeburg, Germany Otto-von-Guericke University, Theory of Complex Networks, Magdeburg, Germany 1 2 Question: One of the most important extracellular mediators of inflammation and immunity is Interleukin 6 (IL-6). One reason for the pleiotropic functions of IL-6 is the complex network of intracellular signalling pathways that are activated by binding of IL-6 to its receptor complex consisting of IL-6Rα and gp130. However, only robust signal transmission allows a cell to appropriately respond to extracellular mediators. Robust signal transmission is perturbed by e.g. varying protein expression levels. In information theory information transmission through noisy channels such as signalling pathways can be quantified by the calculation of the mutual information. Here, we aim to understand robustness in IL-6-induced Jak/STAT signalling by combining multi-colour single cell analysis and computation of mutual information. Methods: To determine mutual information in IL-6-induced signaling the distribution of IL-6Rα, gp130, and STAT3 expression as well as of IL-6-induced phosphorylation of STAT3 were analysed. Ba/F3 cells stably expressing IL-6Rα and gp130 were stimulated with varying amounts of IL-6. Cells were stained with antibodies against IL-6Rα, gp130, STAT3, and phosphorylated STAT3 and subsequently analysed by multi-colour flow cytometry. Additionally, DNA content and cell cycle phase were determined. Based on the multiplexed single cell analysis mutual information between all possible interaction partners is computed to gain a comprehensive map of information transmission within IL-6-induced signalling. Results: IL-6 induces a dose-dependent phosphorylation of STAT3. Of note, the mutual information between the expression of STAT3 and its IL-6-induced phosphorylation is low. This indicates, that the strength of STAT3 phosphorylation is robust and independent of STAT3 expression, in contrast to what would be expected from a simple model of a phosphorylation cycle. This robustness probably enables cells to reliably transmit information by different cytokine concentrations albeit varying STAT3 concentrations. Conclusion: Mutual information analysis helps to define robust structures within signalling pathways that enables cells to cope with biochemical noise and varying expression levels. 56 PP 2-04 Refractory states imprinted in the NF- B system regulate encoding of temporal inflammatory signals Boddington, C.1, Adamson, A.1, Jackson, D.1, White, M.1, Paszek, P.1 University of Manchester, Manchester, Great Britain 1 Cells must rapidly decode changing environmental signals to make fate decisions and coordinate acute tissue-level responses. Robust signal processing strategies must reproducibly discriminate between different cues and regulate appropriate gene expression programs. Here we used time-lapse imaging to show dynamic NF- B system responses to pulsatile cytokine stimulation. Single, or well-spaced pulses of TNFα (>100 min apart) gave a high probability of NF- B activation. However, fewer cells responded to shorter pulse intervals (<100 min) suggesting a refractory state induced by initial stimulation. This refractory state appeared to be imprinted in individual cells, but heterogeneous in the population. Mathematical modelling predicted that it was governed through multiple cellular states or levels of signalling molecules related to the TNFa signal transduction pathway. Such encoding enabled robust and reproducible single cell responses and maintained acute tissue-level signalling. While unresponsive to TNFa, NF- B system was able to encode additional cytokine inputs. This suggested, that refractory states in the NF- B system might enable robust discrimination of multiple temporal cues and thus substantially increasing signal processing capabilities. We hypothesise that such utilization of refractory states imprinted in signalling networks might be advantageous and common to other cellular systems. PP 2-05 Wnt/Planar cell polarity signaling regulates commitment of intestinal stem cells to the secretory lineage Böttcher A.1, Aliluev A.1, Sterr M.1, Büttner M. 2, Sass S. 2, Irmler M. 3, Beckers J. 3, Theis F. 2, Lickert H.1 Helmholtz Zentrum München, IDR, Garching-Hochbrück, Germany Helmholtz Zentrum München, ICB, Neuherberg, Germany 3 Helmholtz Zentrum München, IEG, Neuherberg, Germany 1 2 Introduction Imbalance in intestinal stem cell (ISC) homeostasis leads to cancer as well as inflammatory and metabolic diseases. Thus, defining ISC hierarchy and identifying niche signals that control stem cell fate will provide cellular and molecular targets for therapy. Several ISC populations have been described according to their marker expression, proliferative behavior and lineage potential. In addition, also differentiated intestinal cell lineages possess stem cell character under specific conditions. The classical stem cell pool of actively cycling Lgr5+ ISCs ensures homeostatic renewal and depends on canonical Wnt signaling. A minor population of quiescent, label-retaining cells (LRCs) is commonly viewed as a reserve stem cell pool but is controversial, in part because we lack markers to properly define this cell population. Recent studies identify quiescent LRCs as a subpopulation of Lgr5+ ISCs committed to the secretory fate that can regain stem cell properties following crypt damage. However, the signals that maintain and activate quiescent stem cells, their relationship to the classical Lgr5+ ISC pool and their contribution to homeostatic renewal remain elusive. Material and Methods To define ISC hierarchy and to understand early intestinal lineage decisions we combine mouse genetics with longitudinal single cell imaging and single cell gene expression analysis. Results Here we establish the Wnt/planar cell polarity (PCP) signaling associated gene Flattop (Fltp) as novel marker for intestinal LRCs and their distinct secretory lineage, enteroendocrine cells and Paneth cells. We find that two out of 15 Lgr5+ ISCs per small intestinal crypt express Fltp and become lineage restricted. A subset of Lgr5+/Fltp+ cells gives rise to Fltp Venus reporter-positive cells that are characterized by a combined stem cell and secretory cell gene signature, label-retention (quiescence), mainly locate at crypt position +4, indicative for quiescent intestinal cells and preferentially give rise to Paneth cells. Strikingly, the expression of Wnt/PCP genes rather than a canonical Wnt gene signature distinctive for actively cycling stem cells characterizes Fltp+ cells. In mice with disturbed Wnt/PCP signaling enteroendocrine lineage formation is affected. Taken together, these findings define Fltp as a marker for intestinal LRCs and connect Wnt/PCP signaling to commitment of ISCs to the enteroendocrine and Paneth cell lineage and cell-cycle exit. Conclusion We provide a novel marker to study quiescent intestinal cells and their lineage in homeostatic and diseased state. Furthermore, manipulation of Wnt/PCP signaling could potentially be of interest for the treatment of metabolic diseases or cancer. 57 Program / Abstracts PP 2-06 T cell immune responses generate diversity through linear cell-fate progression Flossdorf, M.1, Buchholz, V. 2, Kretschmer, L. 2, Graef, P. 2, Busch, D. 2, Höfer, T.1 German Cancer Research Center, Theoretical Systems Biology, Heidelberg, Germany Technische Universität München, Munich, Germany 1 2 Upon infection, naive antigen-specific cytotoxic T cells expand vigorously and give rise to a population of short-lived effector and long-lived memory cells. Conflicting models have been proposed that suggest either of these subsets to be a precursor of the other or attribute their generation to asymmetrically dividing naive cells. To gain insight into the mechanism that underlies T cell diversification we combine stochastic population modeling with large scale model discrimination based on single cell in vivo fate mapping data. We developed a computational framework that efficiently incorporates data on single cell dynamics in addition to population mean dynamics. This resulted in significant improvements in both model discrimination and identifiability. Our framework allows for stochastic differentiation and proliferation decisions of individual cells and incorporates both symmetric and asymmetric cell division. Building on this, we find, first, that asymmetric cell divisions of the activated naive T cells play a negligible role and, second, that phenotypic diversity is instead generated through linear cell-fate progression: Naive cytotoxic T cells give rise to slowly proliferating, long-lived subsets from which rapidly proliferating, short-lived subsets emerge. Critical predictions of this linear differentiation model have been validated in subsequent experiments. Third, we find that recall responses initiated by resting memory T cells recapitulate the primary response. Finally, the mathematical model was utilized to quantify the distinct effects of curtailed antigen presentation and inflammation on memory versus effector T cell development. PP 2-07 Estimating single-cell regulatory heterogeneities from cell populations Fuchs, C.1, Bajikar, S. S. 2, Janes, K. A. 2, Theis, F. J.1 1 2 Helmholtz Zentrum München, Institute of Computational Biology, Neuherberg, Germany University of Virginia, Department of Biomedical Engineering, Charlottesville, United States Cell-to-cell variation in gene regulation occurs in a number of biological contexts, such as development and cancer. Discovering regulatory heterogeneities in an unbiased manner is difficult owing to the population averaging that is required for most global molecular methods. Here, we show that we can infer single-cell regulatory states by mathematically deconvolving global measurements taken as averages from small groups of cells. This averaging-and-deconvolution approach allows us to quantify single-cell regulatory heterogeneities while avoiding the measurement noise of global single-cell techniques. Our method is particularly relevant to solid tissues, where single-cell dissociation and molecular profiling is especially problematic. PP 2-08 Universal time and diffusion pseudotime for single-cell data Haghverdi, L.1, Buettner, M.1, Wolf, A.1, Buettner, F.1, Theis, F. J.1 1 Helmholtz Zentrum Muenchen, ICB, Munich, Germany Single-cell gene expression technologies divide into two main categories; time-lapse microscopy and single-cell snapshot expression data. In the first approach expression of a few genes (usually up to three) is measured in actual time for several single cell trajectories. The latter approach measures the expression of several genes (up to 10^4) for several single cells (up to 10^5) sampled at one or few time points but from several developmental stages of the cellular dynamics. We argue that there exists a common manifold for several single cell trajectories following the same dynamics in the gene expression space. Defining universal time (and diffusion pseudotime) as the arc length along this common manifold (on a diffusion space map of the manifold), we established a new framework to bring measurements from these two experimental frameworks together. We illustrate the method on a toy data set as well as large samples of single cell qPCR and RNAseq / dropseq. 58 PP 2-09 Dynamical Modeling of Clonal Evolution in Primary and Relapsed Follicular Lymphoma Horn, M.1, Kreuz, M.1, Loeffler, M.1 University of Leipzig, Institute for Medical Informatics, Statistics & Epidemiology (IMISE), Leipzig, Germany 1 Introduction: Germinal centers (GCs) provide signals for B cells to become high affinity antigen-detecting cells. In this process, somatic hypermutation (SHM) and clonal selection play a pivotal role, resulting in the accumulation of mutations in IG genes. Malignant transformation of GCs to follicular lymphoma (FL) is associated with genetic lesions. The disease can be controlled but not cured, with a median survival of approximately 10 years. In the majority of patients relapses are observed, in some cases additionally associated with transformation to more aggressive diffuse large B-cell lymphoma. Objective: In order to gain insight into processes underlying clonal evolution and lymphoma relapse, we applied a mathematical modeling approach. We particularly sought to explain qualitatively different types of lymphoma evolution, as observed in FL patients, such as divergent, sequential and no evolution. Materials and Methods: We developed a single-cell based mathematical model of physiological GC reaction to study the dynamics of GC expansion and B cell affinity maturation. Furthermore, we applied our model to the situation of FL. We compared our model results to phylogenetic trees reconstructed from clinical measurements (sequences of IGHV rearrangements) of primary and relapse tumor in FL patients [1]. Results: Based on parameter changes in a single cell, representing malignant transformation, the model is capable of reproducing typical features of lymphoma emergence and development. Specifically, different types of evolution can be explained. We identified distinct model scenarios for the “no evolution” case that predict either total absence of SHM in FL or the existence of extra-follicular niche-like environments in which primary FL cells lie dormant for extended periods of time before entering a secondary GC. For sequential and divergent evolution, the model predicts the SHM machinery to work at a significantly reduced rate during the period until relapse as compared to normal GC. Conclusion: Our model is capable of reproducing both physiological GC reaction and transformation to FL. Different scenarios to explain lymphoma evolution need to be validated experimentally. Our findings also have potential clinical implications on e.g. immune therapeutic approaches. Acknowledgement: We thank Prof Ralf Küppers, University Hospital Essen, for many valuable suggestions. Data have been provided by the HaematoSys Consortium. [1] Loeffler M et al. Genomic and epigenomic co-evolution in follicular lymphomas. Leukemia. 29(2):456-63 (2015). PP 2-10 An Interplay of Determinism and Stochasticity in the Information Flow via STAT3 Huang, X.1, Braun, S.1, Harder, N.1, Vanlier, J.1, Alessandro, L.1, Meichsner, J.1, Börner, K.1, Van der Hoeven, F.1, Grimm D.1, Rohr, K.1, Schilling, M.1 1 DKFZ, Heidelberg, Germany Interleukin 6 (IL-6) induced activation of Signal Transducer and Activator of Transcription 3 (STAT3) is an important signal transduction process in liver regeneration and inflammation. Binding of IL-6 to its receptor initiates a cascade of events leading to the accumulation of STAT3 in nuclei that modulates expression of target genes. It is unknown how differences in IL-6 dose are encoded in the alteration of the dynamics of nuclear STAT3. To enable the measurements of the STAT3 dynamics in primary mouse hepatocytes, an mKate2-STAT3 knock-in reporter mouse model was generated. Hepatocytes were stimulated with 12 different doses of IL-6 ranging from 0 to 500 ng/ml and STAT3 dynamics were recorded with time lapse microscopy. Single nucleus tracks were obtained by an automatic image analysis pipeline and statistical modelling was performed. Statistical analysis indicated that the amplitude of the first pulse of nuclear STAT3 encodes the most information about IL-6 dose. Timing of secondary pulses in nuclear STAT3 presented a broad distribution, suggesting that subsequent pulses are less regulated. The dynamics of nuclear STAT3 in two nuclei from one hepatocyte (twin nuclei) were significantly more similar than those in two random nuclei, suggesting that the cytoplasm constitutes a major source of the variability in nuclear STAT3 dynamics. Twin nuclei lose synchrony over time, indicating an increased role of stochasticity at later time points. Signal transduction from IL-6 to STAT3 is a process subjected to both deterministic regulation by IL-6 dose and influence by stochasticity in the signalling pathway. 59 Program / Abstracts PP 2-11 Dynamic NF- B and E2F interactions control the priority and timing of inflammatory signalling and cell proliferation Jones, N.1, Ankers, J. 2, Awais, R. 2, Boyd, J.1, Ryan, S. 2, Adamson, A.1, Harper, C.1, Bridge, L. 3, Spiller, D.1, Jackson, D.1, Paszek, P.1, See, V. 2, White, M.1 University of Manchester, FLS, Manchester, Great Britain University of Liverpool, Institute of Integrated Biology, Liverpool, Great Britain 3 University of Swansea, Mathematics, Swansea, Great Britain 1 2 One of the most important functions in a cell is the accurate interpretation of environmental signals. This is achieved via dynamic signal transduction systems, through which gene expression is re-programmed in response to specific extracellular cues. How cells resolve multiple simultaneous cues without perturbing essential functions is a growing area of importance. Using physiological gene expression through a BAC expression system, we demonstrate that the dynamics of Nuclear Factor kappa B (NF- B) signalling and the cell cycle are prioritised differently depending on the timing of an inflammatory signal. Furthermore Using a combination of computational analyses, and advanced imaging (FRET, FCS) techniques we show physical and functional interactions between NF- B and the E2 Factor 1 (E2F-1) and E2 Factor 4 (E2F-4) cell cycle regulators. We demonstrate that these interactions modulate the NF-kB response and the timing of Mitosis. Treatment at the G1/S boundary resulted in a delayed cell cycle and synchronous NF- B responses between cells. In S-phase, the NF-kB response was delayed or repressed while cell cycle was unimpeded. We believe this is an example of how signaling systems are integrated to prioritise important cell fate decisions. PP 2-12 Inferring cell ensemble models of heterogeneous cell populations by multi-experiment and multi-data-type fitting Kallenberger, S.1, Fröhlich, F. 2, Theis, F. 2, Eils, R.1, Hasenauer, J. 2 1 2 DKFZ, Division of Theoretical Bioinformatics, Heidelberg, Germany Helmholtz Center Munich, Institute of Computational Biology, Munich, Germany Main objectives: In conventional mathematical models of cellular signal transduction networks cell-to-cell variability remains mostly unattended. Live cell imaging together with the increasing number of fluorescence based molecular sensors, however, allows quantitative studies in a variety of biochemical processes at the single-cell level. Therefore, techniques for modeling heterogeneous cell populations are of growing importance. We developed methods to effectively combine time-resolved single cell, flow cytometry and western blot data together with information on cellular events for parameter estimations in a cell ensemble model. Materials and methods: To estimate means and covariances of single cell parameters, we implemented a sigma-point based method for the approximation of mean and variance of the population behavior. In a cell ensemble model, experimental subsets of cells can be included which were exposed to different treatments but can be assumed to have the same covariances and means of kinetic parameters and initial protein concentrations. To synchronize parameter means and variances across multiple experiments, we introduced stochastically motivated penalization terms. Results: We applied our approach to a model of extrinsic apoptosis. This model describes the activation of initiator caspase enzymes and cleavage of the pro-apoptotic substrate BID to tBID, which irreversibly triggers cell death if a certain threshold concentration is exceeded. Using our approach, we could assess the distribution parameters of tBID thresholds. This facilitates quantitative studies of the effects of chemotherapeutics on tBID threshold distributions. The simultaneous analysis of multiple datasets along with appropriate regularization improved the identifiability of single-cell parameter 60 distributions in cell ensemble models. Finally, we characterized the contribution of subsets of experimental data on the identifiability of model parameters. Conclusion: Our novel toolkit enables an efficient and robust inference of cell ensemble models from multiple data types collected under a variety of conditions. The tools are widely applicable and the results can for instance be used to plan, which combination of experimental techniques is required to render model parameters identifiable. PP 2-13 Identifying mediators of disease co-morbidities by integrating omics data. Kastenmüller, G.1, Zierer, J.1, Spector, T.1, Menni, C.1 1 Institute of Bioinformatics and Systems Biology, München, Germany Many common diseases share genetic and environmental risk factors and thus occur together. Age is one of the strongest risk factors for these diseases, hence facilitating comorbidities. In this study we aim to integrate molecular markers of ageing from four different ‘omics’ datasets with clinical phenotypes to identify mechanisms that underlie age-related diseases and cause comorbidities amongst them. To this end, we analysed data from 510 female participants of the TwinsUK study with epigenomics, transcriptomics, glycomics and metabolomics data available. We selected in total 53 markers from these four datasets that were previously reported to be independently correlated with chronological age with high statistical significance. We combined these molecular markers with 92 clinical phenotypes, reflecting amongst others renal function, lung function, body composition as well as environmental factors such as food intake. We inferred a graphical model from the combined dataset using random forest variable importance to assess multivariate dependencies between variables and stability selection to select significant associations while controlling the FWER, as previously described by Meinshausen & Bühlmann (2010). Our model of ageing related markers and phenotypes consists of 316 edges connecting the 145 nodes, which is considerably sparser than the number of pairwise significant correlations (83.5%). Most of the nodes are connected in one giant component (96 nodes), which in advance consists of seven dense modules that represent different physiological functions, such as renal or liver function. Our model recovers several likely causal relationships of molecular markers with phenotypes such as the association of PDE4 with lung function - that were previously studied in model organisms and RCTs. Additionally, our model suggests that IgG glycosylation mediates the co-morbidity of lung disease with renal disease. We further observe that renal disease co-occurs with obesity due to the mediation by the metabolite urate. With a betweenness centrality of 6% the expression of the hormone oxytocin is another central variable in our model. It mediates the association of body composition with inflammatory IgG glycosylation, suggesting a possible mechanism that promotes chronic low-grade inflammation due to obesity. Our model shows that systems biology methods can help to uncover complex relationships between diseases and molecular markers from observational data while reducing the number of mediated associations. This allows to focus on fewer, potentially causal relationships and might thus facilitate drug discovery and reduce adverse drug effects. PP 2-14 Unraveling signaling dynamic patterns with single cell mass cytometry Kumar, S.1, Lun, X. 2, Bodenmiller, B. 2, Rodriguez Martinez, M.1, Koeppl, H. 3 IBM, Rueschlikon, Switzerland University of Zurich, Zurich, Switzerland 3 Technische Universität Darmstadt , Darmstadt, Germany 1 2 Extracellular cues trigger a cascade of information flow, in which signaling molecules gain new functions that ultimately culminate in a phenotypic cellular response. Cells can control this flow of information by controlling the temporal behavior of their signaling molecules. Understanding the temporal dynamics of these responses requires however collecting and analyzing high-quality time series data, while taking into account the individual cell molecular heterogeneity that can alter responsiveness of signaling networks. Recent developments in single cell technologies enable a deep characterization of individual cell responses to external stimuli. For proteins, mass cytometry enables measurement of over 50 proteins and protein modifications in millions of cells. However, unlike experimental approaches that follow single cells over time, mass cytometry destroys cells in each experiment and thus the reconstruction of temporal patterns requires specialized computational approaches. The graphical lasso is an algorithm for learning the conditional independence structure of a graphical model that has become very popular for the reconstruction of undirected graphs. Several available implementations facilitate fast and efficient model reconstructions. However the graphical lasso assumes time-independent data, and thus it cannot be directly applied to infer signaling dynamics patterns from time-series perturbation data. Here we present a generalization of the graphical lasso method to account for structured variations over time. Our 61 Program / Abstracts methodology assumes that the topological structure of the network remains unchanged through the perturbation experiment, and thus a temporal penalty term that forces the model to change smoothly over consecutive time points can be imposed. Model parameters are selected by implementing a variant of the Bayesian information criterion. We have tested our model against the graphical lasso in an in silico study and have found that our method outperforms the graphical lasso in the problem of inferring networks from noisy, biologically smooth data. We have also applied our model to the analysis of single-cell time series stimulation experiments of the EGFR pathway in breast cancer. We have found that our model recapitulates known interactions of the EGFR pathway and characterizes the dynamical evolution of the signaling response, thus providing an efficient and scalable approach to reconstruct the temporal dynamics of signaling responses. PP 2-15 Too Young to Die: Age Structured Population Models Capture Cell Cycle Dependent Apoptosis from Snapshot Data Kuritz, K.1, Müller, F.1, Pollak, N. 2, Allgöwer, F.1 University of Stuttgart, Institute for Systems Theory and Automatic Control, Stuttgart, Germany University of Stuttgart, Institute of Cell Biology and Immunology, Stuttgart, Germany 1 2 Cyclic processes, in particular the cell cycle, are of great importance in cell biology. Continued improvement in cell population analysis methods like fluorescence microscopy, flow cytometry, CyTOF or single-cell omics made mathematical methods based on ergodic principles a powerful tool in studying these processes. Ergodic analysis methods (EA) require the cell population to be in a quasi-steady state, meaning that the proportion of cells in each stage of the cell cycle is stable. This restricts the application of cell cycle analysis with EA to unperturbed cell populations where no cell cycle arrest or cell death occurs. A simultaneously exploration of intracellular signaling and related phenotypes has not been not possible. By relating EA with age structured population models we overcome this restriction and provide a framework to study cell populations that deviate from their steady state distribution. We demonstrate the capability of this approach in a study of cell cycle dependent susceptibility to TRAIL induced apoptotic cell death. Single cell progression through the cell cycle is therefore modeled by a stochastic differential equation on a one dimensional manifold in the high dimensional dataspace of cell cycle markers. Given the assumption that the initial cell population is in a steady state, we derive transformation rules which transform the number density on the manifold to the steady state number density of age structured population models. This transformation is applied to a cell population under TRAIL treatment to study cell cycle dependent caspase cleavage. Furthermore, the changing age structure of the cell population is analyzed with age structured population models to estimate a cell cycle dependent death rate. The estimated death rate demonstrates a higher susceptibility to TRAIL induced apoptosis for cells which are at the end of the G1 phase. This observation is in line with the observed cell cycle dependent initial caspase cleavage. Ergodic analysis can in general be applied to every process that exhibits a steady state distribution. By combining EA with structured population models we provide the theoretic framework for extensions of ergodic principles to distributions that deviate from their steady state. Applications range from studying cell cycle related processes to stem cell development. PP 2-16 GenSSI: Generating Series Structural Identifiability on new Matlab versions Ligon, T.1, Chis, O.- T. 2, Banga, J. 3, Balsa-Canto, E. 3, Fröhlich, F.4,5, Hasenauer, J.4,5 Ludwig-Maximilians-Universität, Physik, München, Germany Institute, Santiago de Compostela, Spain 3 Spanish National Research Council, Vigo, Spain 4 Helmholtz Zentrum, Neuhergberg, Germany 5 Technische Universität München, München, Germany 1 2 Objectives: Ordinary differential equation (ODE) models are widely used to study biochemical reaction networks, but their parameters are often unknown and have to be inferred from experimental data. Structural identifiability analysis tools make it possible to determine which parameters can in principle be uniquely inferred from data, given the dynamics, and the observation and perturbation functions. This analysis can also help to adapt either the model or the experiment for those cases where unique estimation is not possible. However, for most realistic models, these methods are computationally very demanding. Methods: In this study, we consider GenSSI (Generating Series for testing Structural Identifiability), which expands observables in series with respect to time and inputs in such a way that the series coefficients correspond to the successive directional Lie derivatives of the observation function. GenSSI determines global and local structural identifiability of parameters of non-linear ODE models, including those frequently used in systems biology. We have modified GenSSI to use the current version of the Matlab Symbolic Math Toolbox instead of the Maple Toolbox for Matlab. In addition, the code was modified to improve performance, readability, and maintainability. 62 Results: The new version of GenSSI was evaluated using a series of small- and medium-scale ODE models and compared with the results of three alternative tools,EAR, DAISY and COMBOS. The comparison revealed thatEAR and GenSSI work best, but EAR is significantly faster. However, GenSSI provides much more information, distinguishing between local and global identifiability, and shows information about relationships among parameters. Conclusion: The updated version of GenSSI runs on the current MATLAB version (R2015B) and is faster than previous versions (30%). Improved documentation and code are available via GitHub (https://github.com/thomassligon/GenSSI) facilitating its application and further development. PP 2-17 Mechanistic modeling of subpopulation structures for multivariate single-cell data Loos, C.1, Moeller, K. 2, Hucho, T. 2, Hasenauer, J.1 1 2 Institute of Computational Biology, Helmholtz Zentrum München, Neuherberg, Germany University Hospital of Cologne, Department of Anesthesiology and Intensive Care Medicine, Experimental Anesthesiology and Pain Research, Cologne, Germany Main objectives of the study: Cell populations exhibit different degrees of heterogeneity. A better understanding of cellular heterogeneity could advance treatments in a broad range of diseases, since heterogeneity plays an important role for cells ranging from cancer cells to neurons. Accordingly, a holistic understanding of cellular mechanisms requires the investigation of heterogeneity and its sources in the analysis of single-cell snapshot data. For this, models need to capture the subpopulation structures of a cell population and enable the detection of mechanistic differences between the individual subpopulations. Methods: We combine mixture modeling and mechanistic modeling of higher-order moments, which facilitates the detection of differences between subpopulations and experimental conditions and the exploitation of cell-to-cell variability within individual subpopulations. Results: We studied the alteration of subpopulation response by different experimental conditions for multivariate singlecell snapshot data of NGF-induced Erk signaling, a process relevant in pain sensitization. We found that our introduced method provides an improved parameter identifiability and therefore a deeper insight into the mechanisms of the underlying biological system. Conclusion: Our results suggest that our method is a promising tool for the analysis of single-cell data, which might help to achieve a better understanding of heterogeneous cell populations. PP 2-18 Single cell clone transcriptomics derived from murine brown adipose tissue discloses cell type heterogeneity of brown preadipocytes Lutter, D.1,2, Israel, A. 2,3, Ussar, S. 2,3 Institute for Diabetes and Obesity, Helmholtz Zentrum München, Garching, Germany German Center for Diabetes Research (DZD), München-Neuherberg, Germany 3 JRG Adipocytes and Metabolism, Institute for Diabetes and Obesity, Helmholtz Diabetes Center at Helmholtz Zentrum München, München, Germany 1 2 Activation of energy dissipating human brown adipose (BAT) is considered as an attractive therapeutic approach conquering the current epidemic of obesity and its associated morbidities, above all type 2 diabetes. Pharmacological targeting of human BAT is challenged by the the lack of established human BAT cell lines and differences between murine and human BAT. Adult human BAT is a heterogeneous mixture of different adipocyte subtypes, whereas murine BAT is considered as a homogenous fat depot derived from a single precursor. Differentiation of immortalized murine brown preadipocytes, however, suggests heterogeneity also in murine brown adipocytes. To further explore if the morphological and functional differences of the differentiated adipocytes are based upon a heterogenous population of precursor cells, we established 24 clonal cell lines from this immortalized brown adipocyte precursor pool. All clones showed a principle ability to differentiate in vitro, quantified by lipid accumulation. However, we observed differences in the amounts of lipids accumulated as well as the proliferation rates of the precursor cells. Full transcriptome sequencing of these cell lines showed an unexpected high variation in gene expression patterns further underscroing the hypothesis of different adipocyte sub-types. We used laplacian eigenmaps, a nonlinear embedding technique, to discover a low-dimensional projection that reveals eight genecluster, each characterized by an individual gene expression pattern. Differential gene expression of the associated genes revealed a spectrum of overlapping cellular states. Functional analysis of the genes identified distinct biological processes and pathways associated with the individual clusters. Beside identifying cells in different cell cycle states or with activated immune response pathways, we show that different cells activate different metabolic pathways and developmental programs. This unexpected heterogeneity suggests that mouse brown adipocytes, similar to human brown adipocytes, split up into several functional different subtypes defined by distinct precursor populations. 63 Program / Abstracts PP 2-19 Heterogeneity of TLR4 signalling and robust pathogen sensing Paszek P.1 University of Manchester, Faculty of Life Science, Manchester, Great Britain 1 Immune cells exist in our bodies to detect and kill pathogens, but their activation must be controlled to avoid harmful effects. At the population level, activation of toll-like receptor signalling in macrophages, a key pathogen recognition system, results in analogue signal encoding, where average effector response output is fine-tuned by the level of pathogenic signal. However, using single cell transcriptomics and live cell imaging we found that individual cell responses exhibit extreme variability. For example, inflammasome-related IL-1α and IL-1β cytokine shown binary expression patterns such that it is either high or not observed in individual cells. While the response of these cytokines appear random and uncorrelated across the population, a second set of effector signalling molecules display ubiquitous highly correlated expression patterns. These represent distinct regulatory mechanisms of signal integration and generation of cellular heterogeneity via specific signalling networks. Quantitative smFISH analyses suggested that individual cells might have limited capacity to accurately encode different levels of pathogenic inputs. It has often been thought that dynamic biological systems may have evolved to maximize robustness through cell-to-cell coordination and homogeneity. We hypothesize that cellular heterogeneity is an inherent design motif of the inflammatory response that may avoid out-of-control activation and better pathogen sensing in cellular populations. PP 2-20 Cell-to-cell heterogeneity unraveled by computational analysis of single-cell mass cytometry data: cell cycle patterns and trajectories Rapsomaniki, M. A.1, Lun, X. 2, Bodenmiller, B. 2, Rodriguez Martinez, M.1 1 2 IBM Research, Zurich, Rüschlikon, Switzerland University of Zurich, Zurich, Switzerland Advanced single-cell approaches like mass cytometry (CyTOF) allow the simultaneous quantification of dozens of proteins at a single-cell level and, with the use of multiplexing techniques, are able to yield high-dimensional data from a variety of different experimental conditions. When used in cancer research, CyTOF data enable us to study the effect of the inhibition or activation of essential pathways, to characterize inter- and intra-tumor heterogeneity, or to identify cells subpopulations associated with different stages of tumor progression. However, as already demonstrated for the case of single-cell RNAseq data, single-cell technologies are largely influenced by confounding factors, with the cell cycle and the cell volume being the most dominant ones. For instance, unsupervised clustering of mixed populations of single cells reveals that single cells cluster preferentially according to their cell cycle stage instead of their signaling status or cell line. Here, we present a computational approach to account for this hidden source of variability and unravel subpopulations characterized by different signaling fingerprints. By exploiting measurements of 4 established cell cycle markers (IdU, pHH3, Cyclin B1 and pRB), our method initially classifies single cells into discrete cell cycle phases through a prediction step based on decision trees. Then, using a novel trajectory reconstruction technique, information of the 4 markers and the predicted classes is combined to create a new one-dimensional multiplexed continuous marker that represents biological pseudotime. Evaluation of the multiplexed marker values at a single cell level allows the ordering of individual cells on a continuum based on their cell cycle progression. Last, we show how single-cell CyTOF data can be corrected for volume differences using housekeeping gene measurements, and how cell cycle trajectories across different perturbation time points can be used to remove the effect of the cell cycle and reveal signaling patterns. Our proposed method, implemented in a simple and intuitive Graphical User Interface, was tested using CyTOF data from different cell lines (HEK 293T, HCC1937 and THP1) and different types of stimulation (TNF-α and IFN-γ). In summary, our approach allows the deconvolution of cell-cycle and cell volume effects from any type of mass cytometry data, enabling a more accurate classification of cellular populations and analysis of biological processes. 64 PP 2-21 Multiplexed Imaging Cytometry Analysis Toolbox (miCAT) Coupled to Imaging Mass Cytometry (IMC) Reveals Patterns of Cell Interactions Amongst the Heterogeneity of Breast Cancer Schapiro, D.1,2, Jackson, H. 2, Raghuraman, S. 2, Zanotelli, V.1,2, Catena, R. 2, Bodenmiller, B. 2 1 2 Life Science Zurich Graduate School, ETH Zurich and University of Zurich, Zürich, Switzerland University of Zurich, Institute of Molecular Life Sciences, Zürich, Switzerland Background: Current cancer research focuses on the hallmarks of cancer, such as tumor promoting inflammation, escape from immune destruction, and sustained proliferative signaling. Central to many hallmarks is a deregulation of cell-to-cell communication between tumor cells and cells of the tumor microenvironment (TME). An in-depth analysis of the TME requires spatially resolved and highly multiplexed analysis of cell phenotypes and states at the single cell level. Imaging Mass Cytometry (IMC) enables such analyses by imaging of up to 120 proteins and phosphorylation sites simultaneously with sub-cellular resolution. To exploit single cells and their TME for patient stratification, classification and potential biomarker discovery hundreds or even thousands of patient samples need to be measured by IMC and subsequently analyzed by computational tools. Methods: To process, segment, normalize and analyze these (large scale) multiplexed imaging cytometry data, we developed a novel open-source modular, multiplexed imaging cytometry analysis toolbox (miCAT). This user-friendly Matlab based software enables explorative data visualization, analysis modules to study cell types and their marker expression, and several semi- and unsupervised methods for neighborhood analysis within the TME. Finally all of these image cytometry features can be linked with the corresponding clinical data. Results: We used miCAT to process hundreds of IMC images. Our main data set consist of ~200 breast cancer samples including more than 200,000 single cells each with over 50 parameters and dozens of spatial features. The data driven modeling modules implemented in miCAT highlighted a surprising intra- and interpatient TME heterogeneity in all breast cancer subtypes. Furthermore, marker expression, cell types and features of the TME correlated with clinical outcomes. Finally, we found a link between infiltrating CD68+ macrophages and metastatic carcinoma cells based on the expression of markers of the epithelial mesenchymal transition. Validation experiments are currently ongoing. Conclusion: Highly multiplexed imaging approaches, including IMC and serial immunofluorescence imaging, coupled to our computational toolbox are a powerful tool to analyze highly multiplexed single cell tissue data which can be linked to cell phenotypes, morphological features and clinical data to identify stratifying subpopulations and driving forces of metastasis and disease outcome. PP 2-22 Inferring gene regulation using pseudotemporal ordering of single cell snapshots Wolf, F. A.1, Küffner, R.1, Theis, F. J.1 1 Helmholtz Munich, Neuherberg, Germany Question: Can we qualitatively improve the inference of gene regulatory networks (GRNs) when using single-cell instead of bulk snapshots of the transcriptome? Methods: We use a recently proposed diffusion map-based algorithm for pseudotemporal ordering of single-cell snapshots and combine it with different methods for causality analysis of time series. Pseudotemporal ordering refers to solving the “cell ordering problem”, that is to infer - from static snapshot data alone - how cells progress through a certain dynamic biological processes. This allows to reveal hidden dynamic information about the system. Results: We show from first principles that exploiting pseudotemporal ordering allows to substantially exceed the predictive power of state of the art inference algorithms of GRNs. The method obtains a directed GRN, is efficient and scalable, and uses a clearly defined and practically relevant definition of causality. We illustrate the power of the method for simulated and experimental data. Conclusions: The method can be directly applied to single-cell transcriptomic data for developmental processes and we expect it to be widely useful in this case. We will investigate in the future how it can be applied to experiments based on perturbations. 65 Program / Abstracts PP 2-23 Resolving molecular networks and dynamics involved in CD8+ T-cells function and differentiation in acute and chronic infections on a single-cell level Kanev, K.1, Roelli, P.1, Zehn, D.1 Technical University of Munich, Division of Animal Physiology and Immunology, Freising, Germany 1 CD8 T-cells are major players in the adaptive immune defence against intracellular pathogens (including viruses and intracellular bacteria) as well as in tumor surveillance. In the context of acute viral infections like influenza and yellow fever, naive antigen-specific CD8 T-cells differentiate into highly functional effector and memory CD8 T-cells. In result, the generated effector cells swiftly eliminate the viral infections, while the memory cells provide effective protection in case of secondary infection caused by the same virus. In contrast, some viruses like HIV, HBV and HCV in humans as well as LCMV clone 13 in mice are able to establish chronic infections associated with the development of CD8+ T-cells with diminished functional activity, a state often referred as T-cell “exhaustion”. The latter represents a unique state of CD8 T-cell differentiation which is likely to be effective but hyporesponsive, providing the host with the ability to control the viral load without causing severe bystander pathology. Otherwise the excess cytotoxic T-cell activity due to persisting antigen presence can bring the host to even more life-threatening condition compared to chronic viral presence. Some of the general CD8 T-cell hallmarks in different chronic infections compared to acute infection are unique transcriptional program, immune signaling, migration as well as metabolism. Despite the amount of accumulated knowledge about CD8 T-cell phenotype and function in chronic infection, we are still far away from establishing successful strategies for therapeutic reactivation of the immune system in order to eradicate established chronic infections. In our opinion, this might be a reflection of the usually applied population-based assessment of CD8 T-cell gene expression profiles, which is prone to generating biased and even false results due to giving only average values. The main question we want to address is which molecular networks and mechanism are involved CD8 T-cell function and differentiation in acute and chronic infections on a single cell level. To address that question we decided to generate gene expression profiles of CD8 T-cells from different acute and chronic infection settings and time points using single-cell mRNA sequencing approach. So far we are optimizing the method and we have started to obtain single-cell gene expression profiles from different settings. Besides the deep single-cell information on molecular networks and mechanism, we hope that the single-cell approach we apply will be informative concerning cellular dynamics and subpopulation taking place in different acute and chronic infections. Finally, the ultimate outcome will be to exploit that knowledge for generating new therapeutic strategies for successful intervention in immune response against problematic viral infections. PP 2-25 Dimensionality Reduction to Networks: Single Cell Sign Consistency Models Enable Identification of Subpopulations with Distinct Signaling Network States in Mammalian Cancer Cells Tritschler, S.1, Schapiro, D.1, Saez-Rodriguez, J. 2, Bodenmiller, B.1, 2 2 University of Zurich, Institute of Molecular Life Sciences, Zürich, Switzerland RWTH University, Joint Research Centre for Computational Biomedicine Hospital, Aachen, Germany Background: Alterations in signaling pathways underlie many hallmark capabilities of cancer cells. Due to the high connectivity of the signaling networks a systems-based approach is needed to unravel how deregulated signaling relates to a cancer phenotype. This requires highly multiplexed analysis of the activity of signaling components. In addition, to be able to differentiate cancer cell subtypes in heterogeneous tumor samples we need a single-cell resolution. Integration of such highdimensional single cell data for comprehensive visualization and modeling of subpopulations and their signaling network states necessitates the development of computational tools. Methods: Sign consistency models can be used to predict signaling network states from phosphoproteomic data. The key idea is to map experimental data onto a signaling network based on literature (prior knowledge network) using rules on signal propagation. We used an established sign consistency modeling tool (IGGY) and developed a novel framework for single cell multiplexed data. The user-friendly Matlab based toolbox integrates automated data processing and discretization, sign consistency modeling (IGGY), single cell network visualization and clustering algorithms. 500 cells per minute with up to 40 simultaneous measured markers can be modeled on a local computer. Results: Our approach was evaluated with synthetic and real data to show robustness and sensitivity. It was then employed to a mass cytometry data set of epidermal growth factor stimulation experiments in human embryonic kidney cells using 36 markers simultaneously. The data set includes overexpression experiments of mutant signaling proteins known to be relevant in cancer. The hypothesis-driven modeling revealed rare and abundant subpopulations with differences in the activity of major signaling pathways. These distinct network states could be linked to cell cycle and other cellular phenotypes. Validation experiments are currently ongoing. Conclusion: Our framework enables the visualization of multiplexed single cell data in the context of major signaling 66 networks and detects rare subpopulations with distinct network states. The approach reduces dimensionality to networks yet conserves single-cell resolution, which allows to browse large scale signaling data, dissect cell-to-cell variability and thereby generate new hypotheses and improve experimental design. PP 3: Signalling Modelling PP 3-01 Predicting T-helper cell differentiation and plasticity using logical modeling and modelchecking Abou-Jaoudé, W.1, Grandclaudon, M. 2, Monteiro, P. T. 3, Chaouiya, C.4, Soumelis, V. 2, Thieffry, D.1 Institut de Biologie de l’Ecole Normale Supérieure (IBENS), Paris, France Institut Curie, Paris, France 3 INESC, Lisbon, Portugal 4 Instituto Gulbenkian de Ciência, Oeiras, Portugal 1 2 Background: T helper (CD4+) lymphocytes play a key role in the regulation of immune responses. Upon presentation of antigens by antigen presenting cells, naive CD4+ T cells differentiate into various T helper (Th) subsets, which secrete distinct sets of cytokines. This differentiation process requires the integration of multiple signals triggering the T cell receptor, co-stimulatory and cytokine receptors. Diverse combinations of these signals lead to the differentiation of naïve Th cells into diverse subsets, including Th1, Th2, Treg and Th17 subtypes. During the last years, experimental studies highlighted the diversity and plasticity of Th lymphocytes, challenging the classical linear view of Th differentiation and raising new questions regarding the mechanisms which underlie the observed diversity and plasticity of Th subsets. Methods: We investigated the mechanisms underlying this diversity and plasticity using dynamical modeling. In particular, logical modeling has proven suitable for the dynamical analysis of large signaling and transcriptional regulatory networks. We have built a comprehensive logical model of the network governing Th differentiation, which integrates 20 signalling pathways, a dozen transcription factors, and about 30 cytokines secreted by Th cells. To decipher the dynamics of such a large model, we combined a formal reduction method preserving salient dynamical properties with an algorithm enabling the identification of all stable states. Finally we applied model checking tools to get further insights into reachability properties between Th subtypes upon changes of environmental cues and thereby assess the plasticity of Th lineages. Results: We have assessed the consistency of our model by comparing its behaviour with published experimental observations on Th differentiation. Our model qualitatively reproduces the polarisation of naïve Th cells into various canonical subsets (Th0, Th1, Th2, Th17, Treg, Th9, Th22, Tfh) for the corresponding conditions. Moreover, our model predicts the existence of various hybrid Th cells, which express combinations of transcription factors and cytokines associated with distinct canonical subsets. Finally, we used our model to analyse the plasticity of canonical subtypes depending on environmental conditions, as well as to delineate specific Th cell reprogramming strategies. This analysis led to the prediction of many Th-subtype conversions, which still need to be assessed experimentally. PP 3-02 Protein abundance of AKT and ERK pathway components governs cell-type-specific regulation of proliferation Adlung, L.1, Kar, S. 2,3,4, Wagner, M.- C.1, She, B.1, Bao, J. 5, Lattermann, S.1, Boerries, M. 5,6,7, Busch, H. 5,6,7, Timmer, J. 8, Schilling, M.1, Höfer, T. 2,3, Klingmüller, U.1 German Cancer Research Center (DKFZ), Systems Biology of Signal Transduction, Heidelberg, Germany German Cancer Research Center (DKFZ), Theoretical Systems Biology, Heidelberg, Germany 3 University of Heidelberg, BioQuant Center, Heidelberg, Germany 4 Indian Institute of Technology, Department of Chemistry, Mumbai, India 5 IMMZ, ALU, Systems Biology of the Cellular Microenvironment Group, Freiburg, Germany 6 German Cancer Consortium (DKTK), Freiburg, Germany 7 German Cancer Research Center (DKFZ), Heidelberg, Germany 8 Center for Biological Signaling Studies (BIOSS), Institute of Physics, Freiburg, Germany 1 2 Both, the PI3K/AKT pathway and the Ras/MEK/ERK pathway, have been implicated in the control of cell proliferation in many different cell types. The key components of these pathways are highly conserved, which suggests that they form generic proliferation-control modules that function in a specialized, cell-type-specific way. However, the integrated regulation of proliferation as a multistep process, involving signal processing, cell growth and cell-cycle progression, is poorly understood. In this study, we wanted to identify the main determinants of cell-type-specific proliferation control to rationalize targeted intervention strategies such as combined inhibitor treatment. By means of quantitative data generation and mathematical modeling, we investigated signaling dynamics, gene expression profiles and proliferation responses to Erythropoietin (Epo) stimulation of primary murine erythroid progenitor cells and a 67 Program / Abstracts lymphoid cell line. We found that cell-type-specific protein abundance patterns cause differential signal flow along the AKT and the ERK pathway. These signaling dynamics can be explained by assuming global, cell-type-independent enzymatic rate constants that are shared between the cell types and protein abundances of signaling components that are cell-type-specific. With measured protein abundances and the estimated enzymatic rate constants, the mathematical model correctly predicted signaling dynamics of a myeloid cell line that has not been used for model calibration. We were able to link the differentially activated AKT and ERK pathways to downstream regulation of cell growth by S6 and the expression of cell cycle genes. Control of proliferation was exerted mainly by rate-limiting protein synthesis for faster cycling cells while slower cell proliferation was controlled at the level of the G1-S transition. With the integrated mathematical model of Epo-induced proliferation we were able to explain cell-type-specific effects of targeted AKT and ERK inhibitors and faithfully predict their combined effects. Our findings suggest that static protein abundance patterns allow inference of signaling and proliferation dynamics for targeted intervention in the case of individualized cancer therapy. PP 3-03 A SYSTEM BASED TIME SERIES ANALYSIS UNRAVELS PROLIFERATION TO DIFFERENTIATION SWITCH IN ERYTHROID PROGENITORS CELLS UPON ERYTHROPOIETIN STIMULATION Andrieux, G.1, Chakraborty, S. 2, Adlung, L. 2, Schilling, M. 2, Klingmuller, U. 2, Busch, H. 3,1,4, Borries, M. 3,1,4 IMMZ, Systems Biology of the Cellular Microenvironment Group, Freiburg, Germany DKFZ, Division of Systems Biology of Signal Transduction, Heidelberg, Germany 3 DKFZ, Heidelberg, Germany 4 DKTK, German Cancer Consortium, Freiburg, Germany 1 2 Main objective(s): Erythropoietin (Epo) is the main hormone controlling the renewal of red blood cells. Following Eporeceptor binding on colony-forming unit-erythroid (CFU-E), proliferation and differentiation phases are involved in an intricate manner. Although early response pathways, such as JAK-STAT or MAPK are well known, the overall understanding of the switch between proliferation and differentiation is still in pursuit. Therefore we investigated the cellular proliferation to differentiation switch after Epo stimulation based on time-resolved transcriptome and proteome data. Materials and Methods: To this aim, we quantified the gene response and protein abundance using DNA microarrays and Super-SILAC mass spectrometry up to 24 and 5 hours, respectively. Assuming a constant translation rate, we predicted and validated the proteome at late time points and employed a mutual information based algorithm (ARACNE) to reconstruct a regulatory network between transcription factors and Epo-regulated genes based on the predicted gene-proteome data. Results: From the transcription factor network we identified a time-sequential regulatory pattern, wherein proliferation and erythropoiesis related regulators switch their time-sequential activity at around 8 hours after Epo stimulation. Early response genes and transcription factors, such as Cdk4/Cdk9 and Stat5a, are involved in pro-cell cycle processes, whereas late response genes and TFs, e.g. Tfrc/HBB-b2 and Gata1, correlate with erythropoiesis. Conclusion: For the first time we accurately identified the timing of the proliferation to differentiation switch based on high throughput kinetic data. We therefore isolated the main genes regulated in CFU-E upon Epo stimulation and inferred the regulatory network. PP 3-04 The sensitivity of oscillatory properties Baum, K.1, Politi, A. 2, Kofahl, B.1, Steuer, R. 3, Wolf, J.1 MDC Berlin, Berlin, Germany EMBL , Heidelberg, Germany 3 Humboldt University, Institute of Theoretical Biology, Berlin, Germany 1 2 Oscillations occur at many different levels of biologic processes, in genetic systems, signaling and as metabolic oscillations. Examples are circadian oscillations, the canonical NF- B-pathway, and calcium signaling. The oscillations differ in particular in the intensity of their response towards environmental changes, the so-called robustness or sensitivity of the particular response. For example, the period of calcium oscillations is known to be very sensitive while the period of circadian rhythms is very robust. We determine which topological principles render the amplitude or period of oscillatory systems robust by performing sensitivity analyses for chain models as prototype oscillators. In order to avoid bias towards the employed kinetic parameter values, we consider many different parameter sets chosen in an approach based on Monte-Carlo random sampling for the analysis. We examine model properties such as type of feedback, mass conservation relations between species and reaction kinetics and deduce structural principles leading to robust oscillatory properties. We additionally perform sensitivity 68 analyses for published mathematical models of calcium oscillations and circadian oscillations for a variety of different parameter sets and explain their period robustness behavior with the help of the structural principles examined for the prototype oscillator models. PP 3-05 Mathematical Modelling Suggests Differential Impact of β-TrCP Paralogues on Wnt/βCatenin Signalling Dynamics Benary, U.1, Kofahl, B.1, Hecht, A. 2, Wolf, J.1 1 2 Max Delbrück Center for Molecular Medicine in the Helmholtz Association, Berlin, Germany Albert-Ludwigs-Universität, Freiburg, Germany The Wnt/β-catenin signalling pathway is involved in the regulation of a multitude of cellular processes by controlling the concentration of the transcriptional regulator β-catenin. Proteasomal degradation of β-catenin is mediated by two β-transducin repeat-containing protein (β-TrCP) paralogues, homologous to Slimb protein (HOS) and F-box/WD repeatcontaining protein 1A (FWD1), which are functionally interchangeable and thereby considered to function redundantly in the pathway. HOS and FWD1 are both regulated by Wnt/β-catenin signalling, albeit in opposite directions, thus establishing interlocked negative and positive feedback loops. The functional relevance of the opposite regulation of HOS and FWD1 by Wnt/β-catenin signalling in conjunction with their redundant activities in proteasomal degradation of β-catenin remains unresolved. Using a detailed ordinary differential equation model, we investigated the specific influence of each individual feedback mechanism and their combination on Wnt/β-catenin signal transduction under wild-type and cancerous conditions. We found that, under wild-type conditions, the signalling dynamics are predominantly affected by the HOS feedback as a result of a higher concentration of HOS than FWD1. Transcriptional up-regulation of FWD1 reduced the impact of the HOS feedback. Due to the opposite regulation of HOS and FWD1 expression by Wnt/β-catenin signalling, the FWD1 feedback may function as a compensation mechanism against aberrant pathway activation that is caused by reduced HOS expression. By contrast, the FWD1 feedback provides no protection against aberrant activation in adenomatous polyposis coli (APC) protein mutant cancer cells. FEBS J. 2015 Mar;282(6):1080-96. PP 3-06 Bringing Systems Biology to Material Science: Construction of an Information Processing Material Engesser, R.1,2, Beyer, H. 2,3, Weber, W. 2,3, Timmer, J.1,2 University of Freiburg, Institute of Physics, Freiburg, Germany University of Freiburg, BIOSS Center for Biological Signalling Studies, Freiburg, Germany 3 University of Freiburg, Faculty of Biology, Freiburg, Germany 1 2 The rapidly emerging field of synthetic biology is based on utilizing design principles from electrical engineering in biological systems. Synthetic biological networks are usually modularly constructed using well-characterized biological building blocks, like switches, sensors or output modules and applying network topologies derived from engineering disciplines. The characterization of these building blocks is done by mathematical modeling approaches like used in systems biology. In this work we devise and exemplify a design concept of how such computational circuits can be transferred to materials sciences. An information-processing material is synthesized by merging materials sciences with synthetic biology and mathematical modeling methods arising from systems biology. We harness design principles and circuit topologies from electrical engineering to functionally wire synthetic biological switches into a polymer material framework. This strategy allowed us to synthesize polymer materials that perceive input signals, process these signals by fundamental computational operations and produce a corresponding output. We exemplify this design concept by the mathematical model-guided synthesis of an information-processing material that detects and counts light pulses and releases output molecules specific to the number of light pulses detected. This system is characterized by a mathematical model which is used to identify optimal parameter regions to obtain the desired systems output. The here-described concept of merging synthetic biology, mathematical modeling and materials sciences according to design principles from electrical engineering provides the scope to synthesize information processing materials for manifold applications in fundamental and applied research. 69 Program / Abstracts PP 3-07 Mathematical modelling of drug-induced receptor internalisation in HER2-positive breast cancer cell-lines Fehling-Kaschek, M.1, Kaschek, D.1, Peckys, D. 2, Reinz, E. 3, de Jonge, N. 2,4, Korf, U. 3, Timmer, J. 5,6,1 Freiburg University, Physics, Freiburg, Germany Saarland University, Saarbrücken, Germany 3 German Cancer Research Center DKFZ, Division of Molecular Genome Analysis, Heidelberg, Germany 4 INM – Leibniz Institute for New Materials, Saarbrücken, Germany 5 Freiburg University, BIOSS Centre for Biological Signalling Studies, Freiburg, Germany 6 Freiburg University, Freiburg Center for Systems Biology (ZBSA), Freiburg, Germany 1 2 About 10-20% of breast cancer tumours over-express the HER2 receptor. Two drugs applied in medical treatment, Trastuzumab and Pertuzumab, directly bind to the HER2 receptor inhibiting cell growth. The goal of this contribution is to gain a better understanding of the mode of action of these antibodies by mathematical modelling based on ordinary differential equations. A central question for the modelling is the effect of drug-induced receptor internalisation. The internalisation of HER2 receptor in drug-treated cells is directly observed by fluorescence microscopy. Data taken for different treatments from the human HER2-positive breast cancer cell-line SKBR3 is used as input to model the internalisation process. A second model is deployed to investigate the drug-induced cell-signalling. It is based on time-resolved phosphorylation data from the SKBR3 cell-line obtained by reverse phase protein arrays. The data reveals that drug-treatment interferes with cell-signalling at different sites of the downstream pathway. Internalised receptors are a promising explanation for the sitespecific inputs, opening the perspective to finally combine the model of receptor trafficking and signalling. PP 3-08 Simulation-based parameter estimation for kinetic data for the Raf-Mek-Erk pathway Fiedler, A.1,2, Raeth, S. 3, Theis, F.1,2, Hausser, A. 3, Hasenauer, J.1,2 Technische Universität München, Mathematics, Garching b. München, Germany Helmholtz Zentrum München, Institute of Computational Biology, Neuherberg, Germany 3 University of Stuttgart, Institute of Cell Biology and Immunology, Stuttgart, Germany 1 2 Introduction: The signalling pathway built around the Raf-Mek-Erk cascade is known to play a crucial part in many cell processes. We want to study the explicit pathway structure in the context of cell cycle progression using mathematical modelling. Methods: To this end we calibrate candidate models with respect to time resolved dose response data and use model selection to identify the most suitable model. As we have to ensure that the model is initially in steady state, we developed a tailored simulation based optimization method that uses the problem structure. Conclusion: In this study we derived a suitable model that can explain the data and performed the parameter analysis with the newly developed simulation based method that promises good convergence results. PP 3-09 Signal transduction analysis reveals key switches in inflammatory signaling during acute myocardial infarction determining disease recurrence Huber, H.1, Veltman, D.1, Laeremans, T.1, Martin, V.1, Janssens, S.1,2, Sinnaeve, P.1,2, Passante, E. 3 KU Leuven, Cardiovascular Sciences, Leuven, Belgium University Hospital Leuven, Cardiolovascular medicine, Leuven, Belgium 3 University of Central Lancashire, School of Biomedical Sciences and Pharmacology, Preston, Great Britain 1 2 Inflammation is a key process during acute myocardial infarction (AMI), clearing cellular debris and initiating post-infarct cardiac healing and scarring. However, uncontrolled and prolonged inflammation during the initial phase of cardiac repair can lead to malignant remodeling of the cardiac intracellular matrix, cardiac dilation, and eventually disease recurrence and death. Inflammasome-related signaling in leucocytes such as monocytes and neutrophils has been recently proposed as crucial molecular process controlling cardiac inflammation during AMI. The process is initiated by a two stage stimulation by cellular debris and by purine receptor activation, and culminates in the paracrine secretion of inflammatory cytokines IL-1β and IL-18. These cytokines then activate further inflammatory cells in the heart to sustain a stable inflammatory response and activate cardiac fibroblasts to initiate cardiac repair in the subsequent phase. We here present an Ordinary Differential Equation-based model that investigates the control mechanisms of inflammasome activation and regulation in monocytes during AMI. We feed this model with gene expression data obtained both from peripheral blood mononuclear cells of patients from our in house clinics, and using data from gene expression omnibus. We propose a nested feed-forward loop mechanisms that prevents premature and sub-threshold monocyte activation during 70 their recruitment according to recent molecular findings by other groups. We further investigate mechanisms that control termination of inflammation by generating a bistable response through paracrine release of interferons in response to Il1β and IL-18. Interestingly, we found that a switch from type I to type II interferon occurs at the end of the inflammatory phase of AMI, this vastly attenuates inflammasome activation, shifting the on-steady state equilibrium to subthresholdinflammation. We propose that failure of executing this shift in equilibrium leads to incomplete termination of the inflammatory response, contributing to malignant cardiac remodeling and disease recurrence. PP 3-10 Investigating and Modulating Apoptosis Sensitisation in Cultured Cardiac Cells after Exposure to Doxorubicin Using Quantitative Biochemistry and Computational Systems Biology Passante, E.1, Huber, H. 2, Kankeu, C. 2, Clarke, K. 2 University of Central Lancashire, School of Pharmacy and Biomedical Sciences, Preston, Great Britain KU Leuven, O&N I Herestraat 49 , Leuven, Belgium 1 2 Doxorubicin (DOX) is an anthracycline used in the treatment of many human cancer. However its clinical utility is restricted due to several cardiac side effects. Indeed, 10 to 15 years after chemotherapy, more than 10% of patients treated with DOX will develop severe cardiac side effects such as dilated cardiomyopathy and congestive heart failure. Increase of apoptosis in mammalian cardiac cells following DOX exposure has been identified to induce a dilative cardiomyopathy phenotype. We therefore perform quantitative biochemistry in combination with computational models based on Ordinary Differential Equations to investigate if DOX-exposure sensitises cardiomyocytes to apoptosis.Firstly we aim to show that progenitors of cardiomyocytes (cardiomyoblasts) gain protection to apoptosis during differentiation and subsequently we will show that DOX pre-exposure ablates such resistance. We posit that this gain and loss of apoptosis resistance during differentiation and DOX exposure can be best understood by analysing quantitative changes of apoptosis protein levels using a cell culture model and studying the interplay of apoptosis proteins by computational systems models. Our computer models will provide biochemically testable molecular hypotheses as to how DOX-induced apoptosis sensitivity can be reverted by gene silencing and pharmacological interventions directed to modify/alter key player in the apoptotic pathway. PP 3-11 Linking mechanistic understanding of cellular signalling in gastric cancer and cellular phenotypes Hug, S.1, Keller, S. 2, Kneissl, J. 2, Luber, B. 2, Theis, F.1, Hasenauer, J.1 1 2 Helmholtz Zentrum München, Institute of Computational Biology, Neuherberg, Germany Technische Universität München, Institute of Pathology, München, Germany Background: Molecular factors play a significant role in determining the outcome in gastric cancer patients in response to targeted therapy. It is known that the overexpression of several proteins, e.g. EGFR and HER2 are associated with tumour growth and drug responsiveness, e.g. to the monoclonal anti-EGFR antibody Cetuximab. For predicting a specific outcome, statistical and/or mechanistic modelling of molecular characteristics and cellular signalling, as well as an analysis of the signalling-phenotype link have to be performed. Method: For the analysis of cellular signalling, we set up a mechanistic differential equation based model of the EGFR signalling pathway and calibrate it with data obtained for gastric cancer cell lines under different treatment conditions in both known Cetuximab responder and non-responder cell lines. The model can simultaneously handle multiple cell types, namely known responder and non-responder cell lines. The available Western blot datasets are then further analysed with statistical methods to identify recurrent deregulation. Furthermore, we include phenotypic data for the two studied cell lines, where motility, invasiveness and proliferation are measured. These phenotypic data are included into the mechanistic model via a regression approach. To find significant influences, the respective regression coefficients are estimated using Laplacian prior distributions, thus enforcing sparsity. Results: Parameter estimation and model selection provide new insights into the impact of different mutations and Cetuximab response. More specifically, a good fit to the complete measurement data can be achieved by including the MET receptor tyrosine kinase as present in the non-responder cell line and acting as a way to bypass EGFR and induce downstream signaling. Furthermore, preliminary results indicate that phenotypic output is most closely determined by the integral activation of EGFR. Conclusion: The link between a mechanistic cellular signalling model and analyses on the phenotypic level could lead to new insights, such as the possible identification of a predictive biomarker for patient stratification. 71 Program / Abstracts PP 3-12 Dynamic pathway modeling of IL-6 signaling to unravel mechanisms of Drug-Induced Liver Injury Jünger, A.1, Vanlier, J. 2, v.d. Stel, W. 1, Huang, X.1, Braun, S.1, Oppelt, A.1, Timmer, J. 2, Klingmüller, U.1 1 2 DKFZ, Systems Biology of Signal Transduction, Heidelberg, Germany Freiburg University, Institute of Physics, Freiburg, Germany Interleukin-6 (IL-6) is a key effector cytokine with a central role in liver regeneration and inflammatory processes. Recent findings also link IL-6 signaling to the clinical condition of Drug-Induced Liver Injury (DILI), which is an adverse drug reaction of a small population of individuals. DILI is the most frequent cause of acute liver failure and a common reason for the withdrawal of pharmaceuticals from the market. Diclofenac, a regularly prescribed nonsteroidal anti-inflammatory drug, is one of the ten most frequent causes of idiosyncratic DILI. Due to the high complexity of the signaling pathway a systems biology approach will be necessary to unravel the possible contribution of diclofenac to DILI. By applying quantitative immunoblotting and qRT-PCR we were able to demonstrate a synergistic drug-cytokine effect in the human hepatocellular carcinoma cell line HepG2: In comparison to IL-6 alone, IL-6 and diclofenac in combination increase the abundance of nuclear phosphorylated STAT3, a central signal transducer of the IL-6 pathway, as well as the expression of downstream target genes such as SOCS3 and acute phase proteins. Intriguingly, STAT3 target genes show different patterns of responsiveness to the combined drug-cytokine treatment. By integrating quantitative data on pathway components and signaling dynamics, our previously established dynamic pathway model of IL-6 signaling, based on mouse experiments, is continuously adapted to human hepatocellular carcinoma cells. With this mathematical model we intend to elucidate and quantify the crosstalk of diclofenac-induced responses on IL-6 signaling. By identifying the putative targets of the compound, an improved quantitative prediction of DILI and therefore a better risk management will be possible. PP 3-13 Auto-correlation of high-precision NFκB oscillation data for dynamic mean population models of TNFα signaling Kaschek, D.1, Oppelt, A. 2, Huppelschoten, S. 3, Klingmüller, U. 2, van de Water, B. 3, Timmer, J.1,4,5 University of Freiburg, Institute of Physics, Freiburg, Germany German Cancer Research Center (DKFZ), Heidelberg, Germany 3 Leiden University, Academic Centre for Drug Research (LACDR), Leiden, Netherlands 4 University of Freiburg, BIOSS Centre for Biological Signalling Studies, Freiburg, Germany 5 University of Freiburg, Freiburg Center for Systems Biology (ZBSA), Freiburg, Germany 1 2 Dynamic mean population models constitute a frequently utilized model class in systems biology. Those models describe cellular processes deterministically by ordinary differential equations (ODEs) and their predictions can be interpreted as the mean behavior of a cell population. Some dynamic systems exhibit the property that variations in the model parameters, e.g. due to biological variability, lead to qualitatively different predictions. The mean prediction is not a solution of the ODE any more. Consequently, fitting the model to a hypothetical noise-less observation of the mean yields non-zero residuals that might be highly correlated. Here we discuss a TNFα model that is calibrated based on densely sampled, precise NF- B oscillation data and sparse immunoblot data for targets like nuclear factor-kappaB inhibitor (IκBα), IκB kinase (IKK) and NF- B phosphorylation. We develop a strategy to estimate the expected covariance structure and correlation length of residuals from repeated timecourse measurements. Judging from the covariance structure, we conclude that the combination of precise and imprecise data leads to over-fitting of the NF- B oscillations. Conversely, the residual correlation can be employed to adjust the weighting or sampling width of the time-series data. Funding: This work was supported by funding from the MIP-DILI project, a European Community grant under the Innovative Medicines Initiative (IMI) Programme (Grant Agreement number 115336). 72 PP 3-14 Case Study for Attractor Detection of Asynchronous Logical Networks Klarner, H.1, Siebert, H.1 1 Freie Universität Berlin, Mathematik und Informatik, Berlin, Germany In this presentation we first recapitulate a model for the bladder tumorigenisis in humans, which was introduced by E. Remy et al. in [1], and then present new results obtained by applying our attractor-detection algorithm to it, a method published in [2]. The bladder tumorigenisis model is based around the E2F-activating transcription factors and their response to EGFR and FGFR3 stimuli as well as to DNA damage. Its construction includes a previously published model as well as literature research, in particular the Reactome database and the Atlas of Cancer Signaling Networks. It predicts different phenotypes by the asymptotic activities of three read-out nodes: Proliferation which is represented by the activities of CyclineE1 and CyclinA, Apoptosis which is represented by TP53 and E2F1 and Growth Arrest which is represented by RB1, BL2 and p21CIP. Mathematically, the model is a discrete, dynamical network that consists of 30 components and 84 interactions together with respective logical statements that formalize processes like synthesis, degradation and phosphorylation. Such networks are capable of producing two types of asymptotic (or long-term) behaviors: steady states, in which the activity levels of all network components are kept at a fixed value, and cyclic attractors in which some components are unsteady and produce sustained oscillations in the network. To associate possible phenotypes with a given model requires the computation of all possible attractors, a computationally challenging problem. Recently we developed an 0-1 optimization problem to predict the attractors of a network and a model-checking-based algorithm to assess the quality of such predictions. We observed that in practice these predictions agree very well with the real attractors. If time permits we will present some additional results regarding the decision process that must implicitly exist in nondeterministic models capable of producing several different attractors. The goal here is to try to understand the dynamics of logical networkd in terms of two opposing regimes: the decision-making regime in which a subnetwork stabilizes its values independently of the remaining network, and the percolation regime in which state transitions are predetermined and caused by previous decisions. The whole process is depicted in an “attractor decision diagram” and we discuss its possible value for answering questions regarding network control. [1] Remy, E., Rebouissou, S., Chaouiya, C., Zinovyev, A., Radvanyi, F., and Calzone, L. (2015). A modeling approach to explain mutually exclusive and co-occurring genetic alterations in bladder tumorigenesis. Cancer research, 75(19), 4042-4052. [2] Klarner, H., and Siebert, H. (2015). Approximating attractors of Boolean networks by iterative CTL model checking. Frontiers in bioengineering and biotechnology, 3. PP 3-15 Towards genome-scale mechanistic models of signal transduction Krantz, M.1 1 Humboldt-Universität zu Berlin, Berlin, Germany We can build, validate and simulate genome scale metabolic models. In contrast, we still have not managed to create comprehensive signal transduction models: Information transfer networks pose new technical and conceptual challenges, which we never needed to tackle in metabolic modelling. Indeed, it is often difficult to directly apply the methods for metabolic networks on signalling networks. In particular, validation and simulation methods based on mass conservation cannot be used. Furthermore, when we model signal transduction network as mass transfer, we need to enumerate the state combinations the network components can assume. This works well for small models, but rapidly leads to a virtual explosion in model states - and/or simplifying assumptions. As neither is desirable in a knowledge database, we wanted to develop a method to map and model signalling networks at the resolution of signalling data. With this in mind, we developed the reaction-contingency (rxncon) language. It reconciles the need for mechanistic accuracy with the scalability required to tackle large networks. Based on this language we developed a toolbox including generation and simulation of Boolean models uniquely defined by the network. We integrated these methods to create the first workflow that enables reconstruction, validation and gap-filling of signal transduction networks. These methods are analogous to those for metabolic networks, but tailored for signal transduction and information flow. We are currently benchmarking these methods by developing large scale regulatory networks in baker’s yeast. However, these methods are organism independent and could also be applied for mammalian systems. 73 Program / Abstracts PP 3-16 Therapeutic Response Modeling Towards Personalized Medicine Li, J.1,2, Mansmann, U.1,2 1 2 Institute for medical informatics, biometry, and epidemiology, Medical Department, Munich, Germany Deutsche Krebsforschungszentrum, Heidelberg, Germany In order to be able to reflect diverse traits of the human tumorigenesis including sustained proliferation, evading growth suppressors, resisting cell death and other malignancies, we construct a genome-scale molecular signaling map (MSM). By integrating gene expression profiles into the MSM, we could understand why the proliferation of different types of cancer cell lines varies by facing multiple drug treatments. For instance, mTor-, WNT- and Hedgehog pathways greatly contribute to the sustained proliferation in overian cancer cells by the combination inhibition treatment of COX-2 with ERBB3, EGFR, FGFR, PDGFR. By investigating the overall survival of AML patients, we have applied the MSM to propose one of the major reasons of resistance of FLT3 inhibition in AML could be the high activity of JAK-STAT signaling pathway combined with the high expression of ERBB family members including EGF, BTC, TGF-alpha, and HB-EGF. Furthermore, the result of the application of MSM also shows that AML patients with high expression of mir-10a, mir-196a and mir-155 tends to have low overall survival when different phosphatases such as PTPN6 and PTPN11 are not expressed or have significant low expression in patients. Moreover, the MSM possesses the potential to identify the treatement-related biomarkers for individuals. Our current study also identify key biomarker miRNAs inclduing mir-181b, mir-20a, mir-21, mir-31, mir-96, mir-224, mir-374 and mir-503 that might play major roles as oncogenes or tumor-suppressor contributing to tumorigenesis. In sum, we have collected different study-daten including Cancer Genome Atlas and tested the MSM in a systematic way for its therapeutic potential. The results are promising. PP 3-17 Analysis of the TRAIL and cisplatin induced apoptosis and MAPK-PI3K/AKT signal transduction pathways in melanoma with a probabilistic Boolean network approach Lucarelli, P.1, del Mistro, G. 2, de Landtsheer, S.1, Trairatphisan, P.1, Kulms, D. 2, Sauter, T.1 University of Luxembourg, Systems Biology, Belvaux, Luxembourg Center of Regenerative Therapies, Experimental Dermatology, Dresden, Germany 1 2 Despite remarkable scientific and clinical efforts in melanoma research, the incidence of malignant melanoma strongly increased over the last decades. While metastatic melanoma is often characterized by mutation of the kinase BRAF, the large number of different mutations represent the biggest challenge in our understanding of the disease. These mutations strongly increase the aggressiveness of the tumor offering a poor prognosis for the patient. The response to conventional therapeutic treatment usually lasts for a few months only, until drug resistance occurs, leading to tumor relapse which often coincides with increased proliferation and migration rates. The goal of this study is to identify the critical mechanisms leading to drug resistance in different melanoma cell lines. For an in-depth analysis of signal transduction pathways that might confer therapy resistance of melanoma, we apply a systems biology approach, which integrates steady-state time-course data into a mathematical model, to identify the optimal drug combination for a successful treatment. A mathematical model analyzing the crosstalk interactions in death ligand TRAIL- and cisplatin-induced apoptotic signal transduction has been established and refined based on newly acquired experimental time-course data (series of steady states). We employ a probabilistic Boolean network (PBN) modeling approach to analyze the molecular interactions and regulations among signaling molecules with minimal parameterization. This modeling approach was applied to identify the influence of targeted molecules via crosstalk interactions between signaling pathways inducing apoptosis or survival at a quantitative scale. Furthermore, a sensitivity analysis was performed to identify the most sensitive molecular targets in the network which render melanoma cells sensitive to apoptosis and thus allow the formulation of new hypothesis for alternative treatment strategies. Currently, we were able to dissect the critical crosstalk mechanisms of the TRAIL and cisplatin signal transduction pathways by a combination of PBN modelling and sensitivity analysis. In addition, we were able to show the influence of the different pathways for each steady-state. In the next step, we aim to identify which drug combinations could be applied to render resistant melanoma cells sensitive towards apoptotic cell death. The newly identified drug targets and combinations will consequently be validated in selected melanoma in vitro models. 74 PP 3-18 Post-transcriptional regulation by sRNA in Synechocystis PCC 6803 Mader, W.1, Malenka, M.1, Georg, J. 2, Hess, W. 2, Timmer, J.1 University of Freiburg, Physics, Freiburg, Germany University of Freiburg, Institute of Biology III, Freiburg, Germany 1 2 Non-coding small RNA (sRNA) are important regulators in prokaryotes, such as cyanobacteria. While mRNA is decoded in the process of protein translation, sRNA enables quick regulation of gene expression. To this end, sRNA either binds to mRNA and regulates translation, or binds to protein, modifying its function. For the cyanobacterium Synechocystis PCC 6803, the sRNA IsaR1 has been found to regulate RNA expression depending on the availability of iron[1]. Iron is necessary for fundamental mechanisms as respiration and photosynthesis. It is thus essential in many living beings. We investigated the iron regulatory system of Synechocystis PCC 6803 in a network containing the RNAs SufC and SufR and their proteins, as well as the two protein-iron complexes SufC-Fe and SufR-Fe, the sRNA IsaR1 and its complex with SufC-RNA, IsaR1-SufC. Two mutants of Synechocystis PCC 6803 were created. In the first mutant, the sRNA IsaR1 can not be expressed. In the second, the expression of IsaR1 can be triggered by copper pipetted to the medium. The RNA of SufB, SufR, and IsaR1 was measured by micro-arrays, the protein of SufB and SufR was measured by mass spectrometry. Data of the wildtype and the two respective mutants was recorded in an iron-depletion condition, and a condition in which cooper was present in the medium. Synechocystis and its mutants were modeled by differential equations. By maximum-likelihood estimation, we determined relative rate-parameters in the model based on the recorded data. Non-identifiabilities in the models were resolved by model reduction based on profile likelihoods. [1] M. A. Hernandez-Prieto, V. Schön, J. Georg, L. Barriera, J. Varela, W. R. Hess, and M. E. Futschik. Iron deprivation in Synechocystis: Inference of pathways, non-coding RNAs, and regulatory elements from comprehensive expression profiling. Genes, Genomics, Genetics 2:1475-1495, 2012. PP 3-19 A new logic of unified classification of intracellular processes Vatlitsov D.1, Mintser O.1 1 Shupyk National Medical Academy of Postgraduate Education, Medical Informatics, Kyiv, Ukraine Introduction: The basis of living systems is the direct passage of the certain reactions. The systematic study of reactions and processes occurring in certain time at a certain organizational level forms the understanding of the life laws. Therefore, the determination of levels of reactions and concept of creation the integrated classification of processes organization provides the opportunity to study the each pool of reactions separately as integrated in the model that involved the holistic approach process. It is impossible without regulation of relevant knowledge. The aim of study: To formulate a new concept of integrative model of processes classification as a basis of systems biology ontology model and conduction of researchs on intracellular processes regulation. Methods: Were used a content analysis and collocate analysis also were used the data from the databases STITCH, STRING and PANTHER. Results: Were suggested a concept of integrative case organization model that initiate the triggers with 9 different levels of interactions: quantum; ions; molecular; macromolecular; the genetic basis; supramolecular; structural elements of the cell; macrostructural; intercellular. This was the base for understanding the principles of the certain processes compilation in the pools. But the next phase of the present research had defined a problem of interlevels interactions classification. Thus we had proposed universal identifiers for each processes even interlevels interaction. The first step of unified classification is the understanding of the type of processes: automated (which runs without the involvement of intracellular energy) and energydependend processes. Each type of process could be classified at scalable grade on the principle of the amount of involving energy. This classification based on the knowledge that each substance represented as the thermodynamic system at Standard conditions which characterized by a certain margin of internal energy. It could be detect only the changes of internal energy ΔU during the changing of matter in chemical reaction. And the changes of the internal energy involved in reaction at constant volume is the internal energy of the reaction. Conclussions: Was formulated the unified integrative model of living systems based on classification of processes that based on the number and type of reactions’ internal energy and its hierarchical position in the levels of cases interactions for live self-organized systems. 75 Program / Abstracts PP 3-20 Dynamic pathway modelling of TNFα signalling to unravel mechanisms of Drug Induced Liver Injury Oppelt, A.1, Kaschek, D. 2, Huppelschoten, S. 3, Malkusch, S.4, Dörffler, V.4, Herrmann, F.1, Merkt, B. 2, Sison-Young, R. 5, Zhang, F. 5, van de Water, B. 3, Heilemann, M.4 Goldring, C. E. 5 Timmer, J. 2, Klingmüller, U.1 DKFZ, Systems Biology of Signal Transduction, Heidelberg, Germany University of Freiburg, Institute of Physics, Freiburg, Germany 3 Leiden University, Leiden Academic Centre for Drug Research (LACDR), Leiden, Netherlands 4 Johann Wolfgang Goethe-University, Institute of Physical and Theoretical Chemistry, Freiburg, Germany 5 University of Liverpool, MRC Centre for Drug Safety Science, Department of Molecular and Clinical Pharmacology, Liverpool, Great Britain 1 2 Deciphering drug induced liver injury (DILI) not only needs to consider toxic effects caused by a compound on hepatocytes, but also the impact of synergistic drug-cytokine interactions. Studies on diclofenac, a drug ranked among the top 10 causes of DILI, showed that the combined treatment with TNFα caused an increase in apoptosis in vitro. However, the mechanism of how TNFα enhances diclofenac-induced toxic effects remains elusive. Dynamic pathway modelling is a powerful approach to investigate these complex and multifactorial interactions. Quantitative immunoblotting and live-cell imaging was performed in HepG2 cells and primary human hepatocytes upon TNFα treatment and in combination with diclofenac. Subsequently, we developed and validated an integrated dynamic pathway model based on ordinary differential equations (ODEs), which combines the densely sampled live-cell imaging data with the rather sparse immunoblotting data. Our studies reveal immediate, early effects of diclofenac on the TNFα-induced NF- B signalling pathway. Nuclear translocation of NF- B is delayed and the phosphorylation levels of several NF- B signalling components are affected upon the combined treatment of TNFα with diclofenac. Experimental results as well as modelling simulations indicate that diclofenac not only has one single target but is rather affecting several components of the TNF receptor complex and the NF- B signal transduction cascade. Dynamic pathway modelling is used to systematically approach these complex processes leading to enhanced hepatotoxicity. Our aim is to improve the understanding of how drug toxicity interacts with inflammatory signalling to eventually contribute to better test systems for DILI in humans. PP 3-21 Autocrine TGF-beta/ZEB/microRNA-200 signal transduction drives epithelialmesenchymal transition: Kinetic models predict minimal drug dose to inhibit metastasis Rateitschak, K.1, Kaderali, L.1, Wolkenhauer, O. 2, Jaster, R. 3 University Medicine Greifswald, Bioinformatics, Greifswald, Germany University of Rostock, Systems Biology and Bioinformatics, Rostock, Germany 3 University Medical Center Rostock, Gastroenterology, Rostock, Germany 1 2 Question: The epithelial-mesenchymal transition (EMT) is the crucial step that cancer cells must pass before they can undergo metastasis. We aim to develop mathematical models describing the complex changes in signal transduction driving epithelial-mesenchymal transition and how signal transduction can be influenced by drugs that should inhibit metastasis. Methods: In this work we develop four kinetic models that are based on experimental data and hypotheses describing how autocrine transforming growth factor-beta (TGF-beta) signal transduction induces and maintains an EMT by upregulating the TFs ZEB1 and ZEB2 which repress the expression of the miR-200b/c family members. Results: After successful model calibration we validate our models by predicting requirements for the maintenance of the mesenchymal steady state which agree with experimental data. Finally we apply our validated kinetic models for the design of experiments in cancer therapy. We demonstrate how steady state properties of the kinetic models and data from tumorderived cell lines of individual patients can be combined to set up a calibration curve such that the minimal amount of an inhibitor to induce a mesenchymal-epithelial transition (MET) can be determined. Conclusions: We show that kinetic models are able to describe complex molecular biology experiments to study the EMT. Our approach to predict the minimal amount of an inhibitor to induce a MET could be of significant interest for optimizing treatment protocols in the clinics. 76 PP 3-22 Tailored Steady-State Constraints for Parameter Estimation in Non-linear Ordinary Differential Equation Models Rosenblatt, M.1, Timmer, J.1,2,3, Kaschek, D.1 University of Freiburg, Institute of Physics, Freiburg, Germany BIOSS Centre for Biological Signalling Studies, Freiburg, Germany 3 Freiburg Center for Systems Biology (ZBSA), Freiburg, Germany 1 2 Signaling pathways and chemical reaction networks in systems biology are frequently modeled by ordinary differential equations. Model parameters are often unknown and have to be estimated from experimental data, e.g. by maximumlikelihood estimation. To reduce the dimensionality of the parameter space, steady-state information is incorporated in the parameter estimation process where possible. For non-linear models, analytical steady-state calculation typically leads to higher-order polynomial equations for which closed-form solutions are hard to obtain. Even if available, the solutions are not unique and several steady-state expressions have to be considered. This can be circumvented by solving the steady-state equations for kinetic parameters which leads to a linear equation system with a unique solution. When solved for kinetic parameters, however, steady-state constraints tend to become negative for particular model specifications, thus, inducing optimization aborts due to divergent model trajectories. Here, we present an algorithm based on graph theory that derives non-negative, analytical steady-state expressions by stepwise removal of cyclic dependencies between dynamical variables. Our method is applicable to most common classes of biochemical reaction networks comprising mass-action, Hill-type kinetic, inhibition and catalyzation. We show that our approach is especially well-tailored to guarantee a high success rate for optimization. PP 3-23 Dynamics of IL-6-induced classic and trans-signaling Rummel, H.1, Billing, U.1, Christen, H.1, Schliemann-Bullinger, M. 2, Rudolph, N. 2, Bullinger, E. 2, Dittrich, A.1, Findeisen, R. 2, Schaper, F.1 Otto-von-Guericke-University Magdeburg, Department of Systems Biology, Magdeburg, Germany Otto-von-Guericke-University Magdeburg, Institute for Automation Engineering, Magdeburg, Germany 1 2 Main objective of the study: Interleukin-6 (IL-6) is one of the most important inflammatory mediators and currently the target of several therapeutic strategies for the treatment of inflammatory diseases. Two different signaling modes of IL-6 that differ in their biological outcome have been described. Classic signaling via the membrane bound IL-6 receptor alpha (IL-6Rα) acts primarily anti-inflammatory whereas trans-signaling via the soluble IL-6Rα (sIL-6Rα) is responsible for the pro-inflammatory activities. In the past, different strength and dynamics of classic- and trans-signaling have been proposed to be responsible for these different physiological outcomes. Here, we test this hypothesis by analyzing possible differences in the dynamics of signaling induced by classic and trans-signaling. Based on these experimental data a dynamic mathematical model of IL-6-induced classic- and trans-signaling will be developed. Materials and methods: Ba/F3 cells stably expressing glycoprotein 130 (gp130) either with or without IL-6Rα serve as a model for classic and trans-signaling, respectively. Dynamics, dose-dependency and variation of classic- and trans-signalinginduced Jak/STAT- and MAPK-signaling were extensively analyzed by semi-quantitative western blotting and multi-color flow cytometry. Additionally, the influence of multiple inhibitors of classic and trans-signaling-induced signal transduction on cell growth was measured. Results: In contrast to the general idea that IL-6-induced pro-inflammatory trans-signaling is stronger than antiinflammatory classic signaling we found that there are no differences in the strength and dynamics of IL-6-induced signaling between classic and trans-signaling. Additionally, these results are supported by comparable inhibition of the two pathways by thirteen different inhibitors of IL-6-induced Jak/STAT-, MAPK- and PI3K-signaling. Conclusion: Our quantitative data indicates that there are in contrast to former hypotheses no differences in the strength and dynamics of the two signaling modes of IL-6. Therefore, diverse physiological outcomes to IL-6-induced signaling are probably not caused by different signaling dynamics, but more likely by different target cells of classic- and trans-signaling. The results lay the basis for an in depth analysis and mathematical modeling. 77 Program / Abstracts PP 3-24 Selective Control of Upregulated and Downregulated Genes by Temporal Patterns and Doses of Insulin Sano, T.1, Kawata, K.1, Ohno, S.1, Yugi, K.1, Kakuda, H.1 Kubota, H. 2, Uda, S. 2, Fujii, M.1, Kunida, K.1, Hoshino, D.1, Hatano, A.1, Ito, Y.1, Sato, M.1, Suzuki, Y.1, Kuroda, S.1 The University of Tokyo , Tokyo, Japan Kyushu University , Fukuoka, Japan 1 2 Insulin shows postprandial transient secretion with high doses, and fasting sustained secretion with low doses, selectively controlling multiple functions. However, how temporal patterns and doses of insulin selectively control gene expression remains unknown. Here, we analyzed the temporal patterns and doses of insulin-dependent gene expression. We performed transcriptomic analysis of insulin-stimulated hepatoma FAO cells and identified 14 upregulated insulin-responsive genes (IRGs) and 16 downregulated IRGs. The upregulated IRGs responded more quickly to pulse and ramp insulin stimulation, whereas the downregulated IRGs showed higher sensitivity to insulin doses. Mathematical modeling revealed that signaling from insulin to transcription of the downregulated IRGs is more rapid and sensitive to insulin stimulation, whereas transcription of the upregulated IRGs is more rapid. Furthermore, some of the IRGs were consistently upregulated and downregulated by insulin injection in vivo. Thus, transient high-dose insulin selectively regulates the upregulated IRGs, whereas the sustained low-dose insulin regulates the downregulated IRGs. Fig.1. Selective control of the insulin-responsive genes (IRGs) by temporal patterns of insulin. PP 3-25 Sensing and antiviral signaling by RIG-I Schweinoch, D.1, Clausznitzer, D. 2, Frankish, J. 3, Binder, M. 3, Kaderali, L.1 University Medicine Greifswald, Institute for Bioinformatics, Greifswald, Germany Technische Universität Dresden, Institute for Medical Informatics and Biometry, Dresden, Germany 3 German Cancer Research Center (DKFZ), Heidelberg, Germany 1 2 The Retinoic acid inducible gene I (RIG-I) detects intracellular viral RNA (vRNA) and activates the antiviral innate immune response via a mitochondrial signaling pathway. The RIG-I induced signaling cascade triggers the production of type 1 interferons and stimulates antiviral gene expression to interfere with virus infection. Despite intensive efforts, the dynamics of signal transduction within the RIG-I signaling pathway and particularly its regulation are not yet fully understood. Main Objectives: The aim of this study is to develop a mathematical model for the RIG-I pathway using ordinary differential equations (ODEs), which quantitatively describes the innate immune signaling response to RNA virus infection. Thereby, we address the hypothesis of vRNA length dependent activation of the RIG-I pathway and aim at providing insights into the vRNA binding mechanism. Furthermore, we use measurements of the activation of downstream signaling molecules (i.e. interferon regulatory factor 3, IRF3) to analyze the activation dynamics of the mitochondrial adapter protein MAVS, an essential signaling adapter within the RIG-I pathway, which forms protease resistant prion-like aggregates. Materials and methods: The final model will be based on a variety of experimental set-ups. Phosphorylation of intermediate signaling proteins over time has been measured using western blotting techniques, and tracking of pIRF3-eGFP transport into the nucleus has been performed to quantify the rate of transcription factor recruitment. In addition, binding of RIG-I to vRNA of different lengths has been studied by measuring the ATP hydrolysis by the helicase domain of RIG-I upon RNA binding. Moreover, luciferase reporter assays have been used to measure the gene expression induced by the RIG-I pathway for varying vRNA concentrations and lengths. 78 Results & conclusion: Here we present a first model, which is able to reproduce experimental results. Thereby, two independent mechanisms for the initial binding of RIG-I to the vRNA and sequential cooperative attachments of additional RIG-I proteins were able to describe the vRNA binding experiments and are in good agreement with current knowledge. However, additional work will be needed to quantitatively describe the signal transduction mechanisms involved in the RIG-I pathway. PP 3-26 CASPA: A tool for automating the reconstruction of PPI Sub-Networks Sevimoglu, T.1, Arga K. Y.1 1 Marmara University, Bioengineering, Istanbul, Turkey An important objective of systems biology is to enable quantitative prediction of the dynamics of biological pathways. This can be done through the reconstruction of the network structure via the determination of protein-protein interactions (PPI). Nonetheless the ever increasing amount of data provided by genomics and proteomics resources is hard to keep up with. It becomes difficult to analyze and make sense of these data manually most of the time. Consequently the reconstruction of PPI networks with the integration of gene annotation data to select the most appropriate candidate proteins that represent a network is a great challenge. We present here CASPA: a computer aided program to automate Selective Permissibility Algorithm. Our method automates the integration of PPI data with GO annotations to reconstruct a protein interaction network, which is composed of candidate proteins for signal transduction mechanisms in humans and model organisms. The modularity of the workflow allows further extension and additional selection criteria to be incorporated. We have developed a simple user interface where the user provides the selection criterion. CASPA improved protein selection by eliminating human error considering extensive amount of data and resulted in faster candidate protein selection. PP 3-27 One model to rule them all Steiert, B.1, Merkle, R. 2,3, Salopiata, F. 2,3, Depner, S. 2,3, Raue, A.1, Kreutz, C.1, Schelker, M.1, Hass, H.1, Wäsch, M. 2,3, Böhm, M. E. 2, Lehmann, W. D. 2, Schilling, M. 2, Klingmüller, U. 2,3, Timmer, J.1 University of Freiburg, Institute of Physics, Freiburg, Germany German Cancer Research Center (DKFZ), Division Systems Biology of Signal Transduction, Heidelberg, Germany 3 Member of the German Center for Lung Research (DZL), Translational Lung Research Center (TLRC), Heidelberg, Germany 1 2 A major goal in systems biology is to reveal potential drug targets for cancer therapy. Signaling pathways triggering cell-fate decisions are often altered in cancer resulting in uncontrolled proliferation and tumor growth. However, addressing cancerspecific alterations experimentally by investigating each node in the signaling network one after the other is difficult or even not possible at all. Here, we combined quantitative time-resolved data from different cell lines with non-linear modeling under L1 regularization, which is capable of detecting cell-type specific parameters. To adapt the least-squares numerical optimization routine to L1 regularization, sub-gradient strategies as well as truncation of proposed optimization steps were implemented. Likelihood-ratio tests were used to determine the optimal penalization strength resulting in a sparse solution in terms of a minimal number of cell-type specific parameters that is in agreement with the data. The uniqueness of the solution was investigated using the profile likelihood. Based on the minimal set of cell-type specific parameters experiments were designed for improving identifiability and to validate the model. The approach constitutes a general method to infer an overarching model with a minimum number of individual parameters for the particular models. 79 Program / Abstracts PP 3-28 Systematic Analysis of Time-Resolved Transcriptional Signature of the Cross-Talk Between HGF and IL6 Reveals Genetic Program of Hepatocyte Proliferation Control Vlaic, S.1, Hoppe, A. 2, N. Müller, S. 3, D’Alessandro, L. A.4, Braun, S.4, Müller, S.4, Meyer, R.4, Bohl, S.4, Kondofersky, I. 3, Muckenthaler, M. U. 5, Gretz, N.6, Theis, F. J. 3,7, Guthke, R.1, Holzhütter, H.- G. 2, Klingmüller, U.4, Boerries, M.4,8,9, Busch, H.4,8,9 Hans-Knöll-Institute (HKI) Jena, AG Systemsbiology and Bioinformatics, Jena, Germany Charité University Medicine Berlin, Institute for Biochemistry, Berlin, Germany 3 Helmholtz Zentrum München Deutsches Forschungszentrum für Gesundheit und Umwelt, Institute of Computational Biology, Neuherberg, Germany 4 German Cancer Research Center (DKFZ), Heidelberg, Germany 5 University of Heidelberg, Department of Pediatric Oncology, Hematology and Immunology, Molecular Medicine Partnership Unit, Heidelberg, Germany 6 University of Heidelberg, Medical Research Center, Medical Faculty Mannheim, Mannheim, Germany 7 Technische Universität München, Department of Mathematics, Garching, Germany 8 Albert-Ludwigs-University, Institute of Molecular Medicine and Cell Research, Freiburg, Germany 9 German Cancer Consortium (DKTK), Freiburg, Germany 1 2 Liver regeneration is characterized by a sequence of inter- and intra-cellular signaling events starting with an inflammatory phase followed by rapid hepatocyte proliferation until the liver mass is restored. Interleukin 6 (IL6) and hepatocyte growth factor (HGF) play a pivotal role during this process, enhancing and driving proliferation, respectively. While the individual effects of HGF and IL6 have been studied comprehensively, their cross-talk in control of hepatic proliferation is yet largely unknown. Here, we performed time-resolved transcriptome profiling of primary murine hepatocytes stimulated individually or in combination with HGF and IL6 to quantify and model the synergistic actions of these factors on the gene response. A novel statistical approach based on general additive models and systematic gene response categorization selected representative cross-talk genes for inference of a dynamic gene-regulatory network. The in silico model reproduced the time-sequential transcriptional program in the primary hepatocytes initiated by the interplay of HGF and IL6 and was confirmed by independent in vitro experiments as well as in vivo literature data. The analysis highlighted the chemokine Cxcl10 as well as the dehydrogenase Hsd3b4 as putative regulatory hubs in the temporal regulation of proliferation. While both genes are known to be involved in liver regeneration, our study refines their roles within the temporal sequences of events in the early phases of liver regeneration. In conclusion, we showed that a combinatorial stimulus elicits many non-trivial cellular gene responses, highlighting the synergistic nature of gene response regulation. Our combined approach of gene response categorization and subsequent modeling can be applied in general to understand the downstream pathway dynamics arising from combinatorial stimuli in complex biological processes. PP 3-29 Mathematical modeling of the impact of acetaminophen on the HGF-induced cellular responses in primary mouse hepatocytes Vlasov, A.1, Huang, X.1, Schilling, M.1, Klingmüller, U.1 German Cancer Research Center (DKFZ), Division of Systems Biology of Signal Transduction, Heidelberg, Germany 1 Acetaminophen (paracetamol, APAP) is a widely used analgesic compound. However, overdose with APAP can lead to the development of drug-induced liver injury and thereby is one of the most frequent causes of acute liver failure. Up to date, the hepatotoxic effect of APAP is believed to be primarily mediated by its secondary metabolites. However, the molecular mechanisms underlying APAP- induced liver injury remain unresolved. Upon liver injury regenerative responses are activated that involve hepatocyte growth factor (HGF) induced proliferation of hepatocytes and ensure tissue regeneration. We hypothesize that perturbation of HGF-induced signal transduction by APAP could lead to impaired cellular responses and enhanced liver damage. To test this hypothesis we examined the direct effect of APAP on HGF-induced activation of the HGF-receptor Met, the mitogen-activated protein kinase (MAPK) and the phosphatidylinositol 3 kinase (PI3K) signaling pathways and on cell cycle progression in primary mouse hepatocytes. The analysis by quantitative immunoblotting revealed that the activation of both signaling pathways in response to HGF stimulation was reduced upon APAP administration. Furthermore, the extent of HGF-induced DNA synthesis in primary mouse hepatocytes was decreased by APAP in a dose-dependent manner. Live cell imaging followed by single cell analysis of hepatocytes expressing the Fucci2 cell cycle sensor showed that APAP induced a prolonged S/G2-phase, suggesting a G2 arrest. A highly simplified mathematical model of HGF induced signaling based on ordinary differentiation equations (ODEs) was established and calibrated with the obtained quantitative experimental data. This model is currently used to test multiple hypotheses to identify potential intervention points of APAP in the HGF- induced signaling network. It is planned to extend the model with a cell cycle module. Thus, this model will provide new mechanistic insights into regulatory mechanisms controlling hepatocytes proliferation that are impaired during APAP-induced liver injury. 80 PP 3-30 Modeling signaling dynamics with differential equations: are single-cell data and their analysis useful? Wade, J.1,2, Bodenmiller, B. 2, Voit, E.1 Georgia Institute of Technology and Emory University, Wallace H. Coulter Department of Biomedical Engineering, Atlanta, United States University of Zurich, Institute of Molecular Life Sciences, Zurich, Switzerland 1 2 Main objective: Ordinary differential equations (ODEs) are often used to study dynamic cellular processes such as signaling. Classically, experimental data and ODE simulations both represent a population average, yet population averages have also been shown to obscure meaningful biological information. Single-cell experimental methods are continuously increasing to address this issue, and the benefits have been well demonstrated. The purpose of this study is to investigate potential advantages of using single-cell data to model signaling dynamics with ODEs. Materials and methods: We use an ODE representation of a small illustrative signaling system with an artificially created dataset, as well as a small biological dataset, to systematically compare modeling population-average- versus single-cell-type data. Results: Multidimensional single-cell data are empirical distributions and contain information on the relationships among variables in cellular systems (e.g., through correlation). Fitting a model to selected characteristics of the marginal projections of the distribution (or subsets of the distribution) ignores information on this relationship structure. By contrast, one should expect that explicit consideration of the relationship structure improves model fitting. Here, we formally test this expectation. Conclusion: Single-cell measurements represent semi-independent observations of the feasible states of a cellular system. Such information is largely encoded in the complex relationship structure among system variables. Our results show that considering the relationship structure of cellular states revealed by multidimensional single-cell data can be used to improve parameter inference in ODE models of signaling. This result holds independently of subpopulations in the data. PP 4: Metabolism PP 4-01 New Standard Resources for Systems Biology: BiGG Models and Visual Pathway Editing with Escher Dräger, A.1,2, King, Z. A. 2, Lu, J. S. 2, Ebrahim, A. 2, Sonnenschein, N. 3, Miller, P. C. 2, Lerman, J. A.4, Palsson, B. O. 2,5,6, Lewis, N. E.6 University of Tuebingen, Center for Bioinformatics Tuebingen (ZBIT), Tübingen, Germany University of California, San Diego, Bioengineering, La Jolla, CA, USA, United States 3 Technical University of Denmark, Novo Nordisk Foundation Center for Biosustainability, Hørsholm, Denmark, Denmark 4 Total New Energies USA, Inc., Amyris, Inc., Emeryville, CA, USA, United States 5 Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, Lyngby, Denmark, United States 6 University of California, San Diego, Pediatrics, La Jolla, CA, USA, United States 1 2 Background: Genome-scale metabolic network reconstructions enable the simulation and analysis of complex biological networks, thus providing insights into how thousands of genes together influence cell phenotypes. Accuracy in systems biology research requires standards in model construction, a variety of specific software tools, and access to high-quality metabolic networks. Results: To meet these needs, we present the BiGG Models database and a collection of software solutions for model building, curation, visualization, and simulation. BiGG Models contains more than 75 high-quality manually-curated genome-scale metabolic network reconstructions, which can be easily searched and browsed and include interactive pathway map visualizations. These visualizations have been generated with the web-based Escher pathway builder. Escher allows users to draw pathways in a semi-automated way and can visualize data related to genes or proteins that are associated to pathways. An export function facilitates storing Escher maps in the community formats SBML and SBGN-ML. These features make Escher an ideal interactive model development tool. In order to make all models in BiGG Models MIRIAM compliant, BiGG Models itself has become part of the MIRIAM registry and provides links a plethora of external databases for each model component. This rich annotation enables rapid comparison across models. New Systems Biology Ontology terms have been defined that are used to better highlight the role of model components. A comprehensive web API for programmatically accessing the database content enables interfacing with diverse modeling and analysis tools. Conclusions: With these features and tools, BiGG Models provides a valuable database, structured for easy access and to help improve the quality, standardization, and accessibility of all genome-scale models. The development of this resource has boosted the development of community standards for constraint-based modeling. Availability: http://bigg.ucsd.edu, https://escher.github.io 81 Program / Abstracts PP 4-02 Combination of Mass Spectrometry-Based Proteomics and Mathematical Modelling Predicts Therapy Targets of Liver Cancer Erdem, M.1, Egners, A.1, Berndt, N. 2, Mastrobuoni, G. 3, Vvedenskaya, O. 3, Bielow, C. 3, Kempa, S. 3, Holzhütter, H.- G. 2, Cramer ,T.1 Uniklinikum Aachen, Molecular Tumor Biology, General Surgery, Aachen, Germany Charité – Universitätsmedizin Berlin, Biochemistry, Berlin, Germany 3 Max-Delbrück-Center for Molecular Medicine, BIMSB, Berlin, Germany 1 2 Main objectives: Hepatocellular carcinoma (HCC) is becoming a major health problem in western societies due to constantly rising incidence and loss of effective longterm treatment options. Using a combination of proteomics and mathematical modeling we have performed a detailed analysis of the tumor-specific metabolism of HCC in a murine model system (ASV-B mice). Materials and methods: Mass spectrometry-assayed protein abundance values were incorporated into a mathematical model of the central carbon metabolism (CCM) of primary hepatocytes that had been established earlier by the Holzhütter group. This model is based on published kinetic parameters, external metabolite concentrations as well as hormonal (e.g. insulin, glucagon) signaling and enables computational modeling of product generation of various CCM-associated metabolic pathways. Results: The model predicted significant changes of various metabolic pathways in murine HCCs. As expected, glucose uptake and lactate production (the Warburg effect) were higher in tumors compared to benign hepatocytes. In addition, the model predicted lower oxygen consumption and higher mitochondrial membrane potential of cancer cells, arguing for cancerassociated mitochondrial dysfunction. Finally, urea synthesis was predicted to be less effective in cancer cells. Blocking glycolysis with D-glyceraldehyde or 3-bromo-pyruvate resulted in significant inhibition of tumor progression, a functional validation of one model prediction. Furthermore, a significant proportion of mice succumbed to the feeding of a protein-rich diet, validating the predicted loss of urea synthesis efficacy. Conclusions: We present the applicability of mathematical modeling of dynamic metabolic pathways important for tumor biology. Our aim is to model the metabolic response to specific perturbations (e.g. chemotherapy) in order to identify new therapy targets and possibilities for effective combination therapies. PP 4-03 Glucose is not the only source for lactate production by hepatocellular carcinoma cells Haanstra, J.1, Tinneveld, T.1, Merino Tejero, E.1, Bruggeman, F.1, Teusink, B.1 Vrije Universiteit Amsterdam, Systems Bioinformatics, Amsterdam, Netherlands 1 Aberrant signalling and changes to metabolism are both hallmarks of cancer. We aim to understand the interplay between growth factor signalling and metabolism in healthy and cancerous liver cells. For energy metabolism, a well-known phenomenon in cancer is the Warburg effect: the finding that even in the presence of oxygen, lactate is the major end-product of glucose breakdown instead of CO2 (aerobic glycolysis). Lactate production is often used as a measure for aerobic glycolysis. However, even if virtually all glucose ends up in lactate, this does not exclude that other carbon sources like amino acids contribute to lactate formation in cancer cells. Amino acids may be needed for biomass formation but could also fuel parts of the tricarboxylic acid cycle and thereby contribute to ATP synthesis and lactate production. To understand how signalling affects and possibly redirects metabolic fluxes we first need quantitative and time-resolved information on which parts of the metabolic network of hepatocellular carcinoma cells are active under different nutrient conditions. Main objective: Quantification of carbon metabolism in hepatocellular carcinoma cells in the presence and absence of different carbon sources and foetal calf serum (FCS). Materials and methods: We cultivated HepG2 hepatocellular carcinoma cells under different nutrient conditions and followed growth by cell counting. We used HPLC to quantify glucose consumption from the medium and lactate production into the medium during cultivation. Results: Our time-resolved metabolite measurements in the culture medium revealed that HepG2 cells made more lactate than can be explained from the glucose that was consumed. In the absence of glucose as a carbon source, growth of HepG2 cells was only moderately affected and lactate was still produced. Removing glutamine or FCS from the culture medium affected growth more strongly but also under these conditions the glucose that was consumed could not account for all the lactate that was produced. Interestingly, in all these conditions lactate production stopped after some time and was followed by the consumption of lactate. Conclusion: Part of the lactate that is produced by HepG2 cells comes from other sources than glucose, potentially from the amino acids in the medium or FCS. Over time, lactate production stops and the lactate is consumed. Our results show that lactate production alone is not a good measure of aerobic glycolysis in liver cancer cells. 82 PP 4-04 Metabolic profiling of CHO-K1 cells adapted to glutamine-free media Hanscho, M.1,2, Ruckerbauer, D. E.1,2, Galleguillos, S.1,2, Borth, N.1,2, Zanghellini, J.1,2 1 2 Austrian Centre of Industrial Biotechnology, Wien, Austria University of Natural Resources and Life Sciences, Department of Biotechnology, Vienna, Austria Chinese hamster ovary (CHO) cells are preferential production hosts for the pharmaceutical industry due to their ability to produce human like product glycosylation. In order to optimize cell growth and cell viability, growth media are typically supplemented with glutamine. However, excess glutamine leads to high ammonium secretion, a toxic metabolite for CHO cells. Main objectives: Here we aim to analyze, model and understand the response of CHO cells to glutamine supplementation versus glutamine free media based on a multi-omics characterization. Methods: Batch fermentations using CHO-K1 cell lines adapted to grow in glutamine free and glutamine supplemented media provided the samples for transcriptomics, proteomics, metabolomics analyses as well as metabolite concentration profiles of culture supernatants at different time points. These data were analyzed based on the integration into a comprehensive community driven genome scale metabolic reconstruction of CHO cells, called iCHO. Results: iCHO is a highly curated community resource, which combines the knowledge of eight laboratories around the world. Here we show that iCHO was able to correctly predict the growth-rates for the glutamine free and glutamine containing media. These observations allowed us to gain direct insight into the variability of the energy and amino acid metabolism in these differentially adapted CHO cells. Conclusion: iCHO successfully reproduced metabolic alterations upon media modifications and thus highlighted systemwide changes. PP 4-05 The relative importance of kinetic mechanisms and variable enzyme abundances for the regulation of hepatic glucose metabolism - Insights from mathematical modeling Holzhütter, H. - G.1, Berndt, N.1, Bulik, S.1 Charité - University Medicine Berlin, Institute of Biochemistry, Berlin, Germany 1 Background:Adaptation of the cellular metabolism to varying external conditions is brought about by regulated changes in the activity of enzymes and transporters. Hormone-dependent reversible enzyme phosphorylation and changes of reactants and allosteric effectors are the major types of rapid kinetic enzyme regulation whereas on longer time scales also changes in protein abundance may become operative. Here, we used a comprehensive mathematical model of the hepatic glucose metabolism of rat hepatocytes to decipher the relative importance of these different regulatory modes and their mutual interdependencies in the hepatic control of plasma glucose homeostasis. Results:Model simulations reveal significant differences in the capability of liver metabolism to counteract variations of plasma glucose in different physiological settings (starvation, ad libitum nutrient supply, diabetes). Changes in enzyme abundances adjust the metabolic output to the anticipated physiological demand but may turn into a regulatory disadvantage if sudden unexpected changes of the external conditions occur. Allosteric and hormonal control of enzyme activities allow the liver to assume a broad range of metabolic states and may even fully reverse flux changes resulting from changes of enzyme abundances alone. Metabolic control analysis reveals that control of the hepatic glucose metabolism is exerted by only seven enzymes, which are differently controlled by alterations in enzyme abundance, reversible phosphorylation and allosteric effects. Conclusion:In hepatic glucose metabolism, regulation of enzyme activities by changes of reactants, allosteric effects and reversible phosphorylation is equally important as changes in protein abundance of key regulatory enzymes. PP 4-06 Significance test for difference between paired temporal observations Kondofersky, I.1, Erdmann, T.1, Theis, F. J.1, Fuchs, C.1 1 Helmholtz Zentrum, Neuherberg, Germany We introduce a novel statistical significance test for the difference of paired time-resolved observations. A test statistic is constructed similar to a univariate t-test as location, variability and size of the tested data are taken into account. This is done by approximating the time courses with smoothing splines and then calculating and integrating over the functional mean and the functional standard deviation. The formulated test statistic has an unknown distribution, thus for assessing its significance we sample from the null hypothesis with preservation of the functional variability. It is the first statistical test of its kind which is suitable for assessment of global differences in time-resolved and paired data. The developed test is applied on a large number of different artificially created datasets and its dependence on several 83 Program / Abstracts influencing factors such as noise, number of time points per sample, number of samples per group and fraction of missing time points per sample are investigated. Furthermore, the test is compared in terms of receiver operator characteristic (ROC) curves to several other commonly available methods. Finally, the test is applied to real-world data where metabolomics datasets from nutritional challenges are investigated. The results suggest almost no metabolic differences between standardized and non-standardized meals prior to a nutritional challenge and thus provide evidence for unnecessary meal standardization in nutritional pilot studies. PP 4-07 A mathematical model of the bile enterohepatic circulation Żulpo, M.1, Balbus, J.1, Wrona, A.1, Kubica, K.1 Wroclaw University of Technology, Department of Biomedical Engineering, Wroclaw, Poland 1 Objectives:. High blood cholesterol has been identified as a classic coronary risk factor. Low cholesterol balanced diet is recognized as a major prevention route of cardiovascular diseases (CVD). However, the connection between dietary cholesterol and CVD remains ambiguous and short term responses to high cholesterol meals are not fully understood. It is not surprising that disruption of the bile enterohepatic circulation affect the cholesterol homeostasis. Also individual features are crucial in these processes. Majority of patients exhibit elevated blood cholesterol level after meal (hyperresponders) but also significant group exhibit the inverse relationship (hypo-responders). Many processes and disorders demonstrate undeniable role of the gallbladder in enterohepatic cycle and therefore in our three-compartment model we pay a special attention to its modelling. Methology: We have examined cholesterol homeostasis using a mathematical model, which consists of a set of the three differential equations and is solved by Runge-Kutta method. The solutions of the model are time-concentration dependences of cholesterol in the liver, peripheral blood and bile in the gallbladder. Our model allows to account for the synthesis of cholesterol in liver, lipoprotein transport from and to the liver, portal circulation, demand of the body tissue and diet cholesterol. This model allows to analyze the impact of the inhibitors of synthesis of cholesterol, as well as drugs being a bile acid sequestrants. It is also possible to show the effect of increased intestinal peristalsis, abnormal amounts of lipoprotein cell receptors and dietary cholesterol on cholesterol homeostasis. Results: The developed three-compartment model allows to study the influence of bile circulation on cholesterol homeostasis. Basing on the model analysis we have found that: a) the influence of the dietary cholesterol on blood cholesterol level is similar for gallstone and gallstone free patients, b) gallstone patients (GP) are more sensitive to the bile acid sequestrants. which regulate the amount of removed bile. Lower bile level decreases total blood cholesterol level, increases the rate of cholesterol and cholic acid synthesis. However, the ratio of the rate of cholesterol synthesis to the rate of cholic acid synthesis drops, c) inter-individual features of bile circulation might result in different response to the dietary cholesterol (hyper and hyperresponders). Acknowledments: This work was partially supported by a research grant from the National Science Centre (DEC2011/03/B/NZ4/02390) PP 4-08 New insights in the relation of liver and adipose tissue via Hedgehog Signalling Matz-Soja, M.1, Rennert, C.1, Gebhardt, R.1 1 Institute of Biochemistry, Faculty of Medicine, Leipzig University, Leipzig, Germany Hedgehog (Hh) and Wnt/ß-catenin signalling pathways are morphogenic cascades and essential for embryonic development and tissue differentiation. On the other hand, both signalling pathways regulate elementary, metabolic functions in adult tissues, despite a significantly reduced activity compared to embryogenesis. From our previous work, we know that in the adult liver lipid metabolism and the IGF-homeostasis is regulated by the Hh pathway and kept in balance. Interruptions in the activity of the Hh pathway lead to steatotic livers in mice which resembles the picture of non-alcoholic fatty liver disease (NAFLD). In addition, we observed that the modulation of Hh signalling pathway in hepatocytes appears to cause a dramatic effect on the adipose tissue. To investigate the link between Hh signalling in hepatocytes and adipose tissue, various hepatocyte-specific knockout mouse models have been bred for activation and inactivation of this cascade. The inactivation was carried out by deletion of Smoothened (Smo), which is an important receptor protein in the signaling cascade. The Hh activation was achieved by inhibition of the receptor protein Patched 1 (Ptch1). To take in account the influence of nutritional factors, a high fat diet (HFD) was administered for 4 and 10 weeks to the transgenic animals. Subsequently, the various adipose tissue species (brown, subcutaneous and visceral) were collected and used for further analysis. The results of our studies showed that there is a, hitherto completely unknown, relation between the activity of Hh signalling 84 pathway in hepatocytes and adipose tissue. The mice with hepatocyte-specific inactivation of Hh cascade showed a significant increase in white adipose tissue (WAT), whereas the mass of brown adipose tissue (BAT) does not change. In contrast, we observed that a hepatocyte-specific activation of Hh signalling cascade leads to a significantly lower fat deposition in the WAT, after administration of a HFD for 10 weeks. On the transcriptional level, the results showed that the expression of thermogenic genes were altered in the hepatocytes due to the modulation of the Hh signalling pathway. From the present results it can be concluded that the amount of the WAT is inversely correlated with the activity of Hh signaling pathway in hepatocytes. The crucial question is, which Hh-dependent mechanism are used for the liver/adipose tissue communication. Answering this question is the goal of further work. PP 4-09 Why Respirofermentation? Explaining the Warburg effect in tumour (and other) cells by a minimal model Möller, P.1, Boley, D. 2, Kaleta, C. 3, Schuster, S.1 Friedrich Schiller University Jena, Bioinformatics, Jena, Germany University of Minnesota, Computer Science & Engineering, Minneapolis, United States 3 Christian-Albrechts-University Kiel, Research Group Medical Systems Biology, Kiel, Germany 1 2 Main objectives: Tumour cells mainly rely on glycolysis leading to lactate rather than on respiration to produce ATP. This phenomenon is known as the Warburg effect (named after German biochemist Otto Warburg) and also occurs in several other cell types such as striated muscle cells, activated lymphocytes, microglia, and endothelial cells. It seems paradoxical at first sight because the ATP yield of glycolysis is much lower than that of respiration. An obvious explanation would be that glycolysis allows a higher ATP production rate, but the question arises why the organism does not re-allocate protein to the high-yield pathway of respiration. Materials and Methods: We model this by a linear programming problem in which not only the rates but also the maximal velocities are variable and tackle this question by a minimal model only including three combined reactions. We consider the case where the cell can allocate protein on several enzymes in a varying distribution. Results: Here, we present a minimal model for explaining the higher ATP production in the Warburg effect. The model predicts pure respiration, pure fermentation or respirofermentation, depending on protein costs and substrate availability. A reallocation of protein to the high-yield pathway only pays if the synthesis costs for that pathway are low enough. Conclusion: We focused on the explanation of the Warburg effect in terms of ATP production rate. It is worth noting that this explanation is consistent with the argument in terms of precursor supply. Whenever the glycolytic rate is high while respiration is limited, much pyruvate can be used for building biomass. A useful feature of our model is that it can be treated analytically. It is linear in the reaction rates, while the underlying rate laws can be nonlinear. PP 4-10 Context-specific metabolic modelling reveals cell-type specific epigenetic control points of the macrophage metabolic network Pires Pacheco, M.1, Nikos, V. 2, John, E. 3, Pfau, T.1,4, Kaoma, T. 5, Heinäniemi, M.6, Nicot, N. 5, Vallar, L. 5, Bueb, J.- L.1, Sinkkonen, L.1,3, Sauter, T.1 University of Luxemburg, Esch-sur-aLzette, Luxembourg Adobe Research, San Jose, United States 3 Luxembourg Centre for Systems Biomedicine, Esch-sur-Alzette, Luxembourg 4 University of Aberdeen, Aberdeen, Luxembourg 5 Luxembourg Institute of Health, Luxembourg, Luxembourg 6 University of Eastern Finland, Finland, Luxembourg 1 2 Objectives: 1) Enable the reliable building of context-specific algorithms that could potentially be used in a near future as tool for the routinely diagnosis and analysis of metabolic disease. This requires fast algorithms with a high level of predictive power, robustness to noise and sensitivity. 2) Establish benchmarking procedures that allow assigning context-specific algorithms in a standardized and unbiased manner. 3) In the application to the macrophage metabolic network, identify epigenetic control points in the metabolism of macrophages. Material and Methods: FASTCORE (Vlassis et al. 2014) and FASTCORMICS (Pacheco et al. 2015), are devoid of heuristic parameter settings and allows building of context-specific models in the time order of seconds and minutes respectively. FASTCORE requires as input a set of reactions that are known to be expressed in the context of interest (core reactions set) and a genome-wide reconstruction. FASTCORE uses an approximation of the cardinality function to force the core reaction set to carry a flux above a user-defined threshold ε. Then applies L1-minimization to penalize the activation of reaction 85 Program / Abstracts with low confidence level while constraining the set of core reactions to carry a non-negative flux. Whereas FASTCORMICS includes additionally a discretization step based on BARCODE (Zilliox & Irizarry 2007; McCall et al. 2011) to determine active and inactive reactions based on solid statistical evidences, in order to cope with probe effects and noise affecting microarray experiments. The prediction power, robustness and sensitivity of FASTCORE and FASTCORMICS along with 5 competing algorithms were tested within benchmarking procedures using real and artificial data. This was based among others on cross-validations in order to assess how the models varied when 10% to 50% of missing input data and on the assessment of the expression confidence level of the genes included in the model via Barcode z-scores and the Human Protein Atlas and confidence scores (Pacheco et al. 2016). Results and conclusion: FASTCORE and FASTCORMICS were shown to outperform their competitors in speed and in the capacity to capture small metabolic variations between tissues (Vlassis et al. 2014; Pacheco et al. 2015; Pacheco et al. 2016). Further, the FASTCORE family models showed the highest enrichment in genes with high confidence levels. Surprisingly the benchmarking procedures revealed that the models reconstructed by the tested algorithms when fed with the same input shared for the more distinct models only around 30% of the reactions (Pacheco et al. 2016). In an application, the mapping of ChIP-Seq experiments for histone H3 lysine27 acetylation (H3K27ac) on the macrophage metabolic model built by FASTCORMICS via the integration of microarray data showed that genes under high regulatory load have the most cell type-restricted and abundant expression profiles within their respective metabolic pathways and are associated to reactions significantly enriched for transport reactions and other pathway entry points, suggesting that they are the critical regulatory control points for cell type-specific metabolism (Pacheco et al. 2015). PP 4-11 How Hedgehog signaling pathway activity controls steroidogenesis in the liver Rennert, C.1, Matz-Soja, M.1, Gebhardt, R.1 1 Institute of Biochemistry, Faculty of Medicine, Leipzig University, Leipzig, Germany The liver is one of the largest organs in our organism with multiple anabolic and catabolic functions which can be carried out simultaneously only by zonation. Recent studies showed that Hedgehog signaling, a morphogenic pathway commonly associated with embryogenesis, development and cancer is active in adult hepatocytes and acts as master regulator of zonation in the adult liver (Gebhardt & Matz-Soja, WJG, 2014). Our goal is the investigation of the gender dimorphism of gene regulation in the liver. So far, we surprisingly found a regulation of steroidogenesis associated genes in hepatocytes of mice with aberrant Hedgehog signaling; yet it was assumed that steroidogenesis occurs in the liver only during embryogenesis and is down-regulated afterwards. A prominent gene we found is Cyp17a1, a central regulator of steroidogenesis. We generated two transgenic mouse strains with an inactivated and constantly-activated Hedgehog signaling pathway in hepatocytes, respectively. The first mouse strain has a hepatocyte-specific knockout of Smoothened (Smo), whereby the inhibition complex of the GLI transcription factors is continuously active, which results in an inactivated signaling cascade (SAC mice). The second strain has a hepatocyte-specific knockout of Patched1 (Ptch1) which lead to a permanently active Smo receptor whereby the inhibition complex of the GLI factors is inactivated and the signaling cascade is continuously active (Ptch1LC1 mice). We found, that a down-regulated Hedgehog pathway results in the up-regulation of some of the steroidogenic genes and vice versa. Cyp17a1 is up-regulated in SAC knockout mice and decreased in expression in Ptch1LC1 knockouts in comparison to wild types. Furthermore we observed an analogous regulation of estrogen receptor (Esr1) expression, relevant for steroid hormone mediated signaling, with a clear correlation to the regulation of Cyp17a1 expression in both transgenic mice strains. A transcription factor binding site analysis revealed GLI binding sites in the promotor region of Esr1. For Cyp17a1 those binding sites were not found. It needs to be further investigated how the regulation of steroidogenesis is influenced by Hedgehog signaling and how Cyp17a1 and Esr1 expression is regulated by each other. Collectively, the experiments showed a clear influence of the morphogenic Hedgehog pathway on the regulation of steroidogenesis in the liver. These unexpected findings are promising for future studies to improve our understanding of the gender dimorphism of regulatory mechanisms in liver. 86 PP 4-12 Cause and Cure of Sloppiness in Ordinary Differential Equation Models Tönsing C.1, Timmer J.1, Kreutz C.1 University of Freiburg, Institute of Physics, Freiburg, Germany 1 For the purpose of mathematical modeling of biochemical reaction networks by the frequently utilized nonlinear ordinary differential equation (ODE) models, parameter estimation and uncertainty analysis is a major task. In this context the term sloppiness has been introduced recently for an unexpected characteristic of nonlinear ODE models. In particular, a broadened eigenvalue spectrum of the Hessian matrix of the objective function covering orders of magnitudes is observed, although no such hierarchy of parameter uncertainties was expected a priori. In this work, it is shown that sloppiness originates from structures in the sensitivity matrix arising from the properties of the model topology and the experimental design. It will be clarified that the intensity of the sloppiness effect is controlled by the design of experiments, i.e., by the data. Thus, we conclude that the assignment of sloppiness to a model as a general characteristic is incomplete without discussing experimental design aspects. Furthermore, we validate this proposition by presenting strategies using optimal experimental design methods in order to circumvent the sloppiness issue and show results of non-sloppy designs for a benchmark model. Reference: Tönsing C., Timmer J., Kreutz C., Cause and cure of sloppiness in ordinary differential equation models, Phys. Rev. E 90, 023303 (2014) PP 4-13 Sexual dimorphism during development of cyp51 liver conditional knockout mice Urlep, Z.1, Lorbek, G.1, Juvan, P.1, Perse, M. 2, Jeruc, J. 2, Björkhem, I. 3, Matz-Soja, M.4, Gebhardt, R.4, Rozman, D.1 Medical Faculty, University of Ljubljana, CFGBC, Institute of Biochemistry, Ljubljana, Slovenia Medical Faculty, University of Ljubljana, Medical Experimental Centre, Institute of Pathology, Ljubljana, Slovenia 3 Karolinska Institute, Karolinska University Hospital, Huddinge, Department of Laboratory Medicine, Division of Clinical Chemistry, Huddinge, Sweden 4 Medical Faculty, University of Leipzig, Institute of Biochemistry, Leipzig, Germany 1 2 Main objectives: With the advancement in personalized medicine, differences between sexes have been gaining increasing attention. Liver is a tissue with one of the highest levels of sexual dimorphisms. Studies focusing on NAFLD, NASH and HCC show male predominance, although the results can vary with respect to age, race and type of disease (e.g. lean NASH). Cholesterol is inherently important for cell function and survival. Blocking its synthesis in the liver causes severe histological and metabolic changes. In the present study we examined the role of sex during development of Cyp51 liver knockout (LKO) mice with a defect in hepatocyte cholesterol synthesis. Methods: From 704 mice of different genotypes, 46 died or had to be euthanized between 4 - 10 weeks of age. All were of the KO genotype, termed runts (KOR). They were evaluated histologically, biochemically and genetically in order to assess the phenotype, developmental stage and changes to the liver. Affymetrix microarrays were used to evaluate global expression with special attention on sexual dimorphism. Results: Few differences separate sexes in Cyp51 liver KOs during development, however the emergence of KO runts after 4 weeks is clearly male predominant with a 2:1 ratio. Histologically, there is no difference between sexes in KO runts, but microarrays revealed 23 differentially expressed genes. Upregulation of caspase 12 (Casp12) points to male susceptibility towards ER stress. KEGG pathway enrichment revealed better conservation of metabolism in KOR females, while inflammatory processes were stronger in males. Similarly, male enrichment of NFY transcription factors, which play a role in unfolded protein response, points to female-based stress protection. Surprisingly, the results in adult KOs contrast the observations in KORs, with females exhibiting decreased metabolism and increased inflammation. Conclusion: In light of increasing gender awareness in liver pathologies, we show that liver cholesterol ablation by Cyp51 knockout progresses differently in males and females. During development, the female gender seems to offer some stress protection, however the higher requirements for cholesterol in adult females may put them to greater risk thereafter. 87 Program / Abstracts PP 5: Multi-Scale Approaches PP 5-01 Tissue architecture representation in pharmacological models. Insights from liver. Boissier, N.1, Celliere, G.1, Friebel, A. 2, Hoehme, S. 2, Hengstler, J. 3, Vignon-Clementel, I.1, Drasdo, D.1 Sorbonne Universités, Inria, UPMC Univ Paris 06, Lab. J.L. Lions, PARIS, France University of Leipzig, Interdisciplinary Centre for BioInformatics, LEIPZIG, Germany 3 IfADo, Systemtoxikologie, DORTMUND, Germany 1 2 Question: Mathematical models of drug clearance and organ metabolism are increasingly used to obtain quantitative predictions of drug kinetics or metabolite levels in the body. The tissue architecture therefore needs to be taken into account. Yet, its influence on tissue function remains unclear. We propose to study the impact of the tissue architecture representation on the predictions obtained from pharmacokinetic models. Because liver is the main detoxifying organ, we have taken the liver tissue architecture as the example for this study. Methods: We compared steady-state outlet concentrations as well as dispersion induced by the space inhomogeneities for several spatial models, from one or several compartments, different spatial alignment of cells along blood vessels to an imagebased representative liver lobule (anatomical unit of liver). In the blood vessel network, blood is assumed an incompressible Newtonian fluid in a series of connected cylinders that flows according to Poiseuille’s law and mass conservation at junctions. The variation of apparent viscosity with diameter is following Pries’ law derived for in-vivo conditions [1]. Transport of the compound in the vasculature is then modeled and coupled with the metabolism inside the cells. The representative liver lobule, generated by Hoehme et al [2], is based on geometrical statistics extracted from confocal image analysis using TiQuant [3]. Results and Conclusions: The simulations reveal that the impact of the precise representation of space depends on the order of the reaction kinetics. A space-discretization along the direction of the compound concentration gradient leads to a higher extraction of the compound whereas considering the flow inhomogeneities leads to a lower extraction. We show that the choice of how space will be taken into account in the model is crucial when the results are used quantitatively, to explain data or to predict drug effects. [1] Pries A. R., Secomb T. W., Gaehtgens P., “Biophysical aspects of blood flow in the microvasculature.” Cardiovascular research. , 32(4), 654-667 (1996). [2] Hoehme S., Brulport M., et al. “Prediction and validation of cell alignment along microvessels as order principle to restore tissue architecture in liver regeneration.” PNAS 107.23 10371-10376 (2010) [3] Hammad S., Hoehme S., Friebel A., et. al. “Protocols for staining of bile canalicular and sinusoidal networks of human, mouse and pig livers, three-dimensional reconstruction and quantification of tissue microarchitecture by image processing and analysis.” Archives of Toxicology 88 (5) 1161-1183 (2014) PP 5-02 Agent-based modelling characterises the effect of localized versus spread damage among mitochondrial population Dalmasso, G.1, Hamacher-Brady, A.1 German Cancer Research Centre (DKFZ), Lysosomal Systems Biology, Heidelberg, Germany 1 When a cell is subjected to metabolic or environmental stresses, mitochondrial fusion and fission are fundamental in conserving a population of healthy and operative mitochondria. In particular, fusion reduces cellular stress rescuing partially dysfunctional mitochondria by redistributing damaged constituent, whereas fission allows the segregation and removal of damaged parts. Moreover, fission is essential in the generation of new mitochondria. A fine balance between fusion and fission machineries is therefore needed for cell survival, and disorders in these processes can lead to neurodegenerative diseases, e.g. Parkinson’s. In order to investigate how mitochondrial damage affects cellular and mitochondrial population homeostasis, here we present a systems biology approach, using agent-based modeling (ABM), to investigate the link between mitochondrial damage and stress response within single cells. Specifically, we modeled a population of fusion-and-fission capable mitochondria which can undergo different damage levels and are able, accordingly, to discard damaged parts or being rescued by fusion. Moreover, in order to profit from the intrinsic spatial component of ABM we analyzed the effect of localized versus spread damage among mitochondrial population. As a result we efficiently (1) discerned which is the natural emerging damage threshold the cell is able to handle in localized versus spread damage scenarios, (2) estimated the role of degradation, fusion, fission and biogenesis on the cellular damage response, and (3) determined the maximum amount of stress a cell can handle comparing fixed and dynamic probabilities of degradation, fusion, fission and biogenesis. This modeling approach represents a novel framework to combine in vitro and in vivo experimental results and could be used to bring new insight in understanding the link between mitochondrial dynamics and cell behavior, crucial requisite for developing new treatments strategies of mitochondrial diseases. 88 PP 5-03 Morpheus 2.0: an open-source framework for multi-scale multicellular systems biology de Back, W.1,2, Starruß, J. 2, Brusch, L. 2, Deutsch, A. 2 TU Dresden, Faculty of Medicine “Carl Gustav Carus”, Dresden, Germany TU Dresden, Center for Information Services and High Performance Computing, Dresden, Germany 1 2 Morpheus is a modeling and simulation environment for multicellular systems biology that provides tools for the simulation and integration of differential equations, reaction-diffusion systems and cell-based models. It facilitates the creation of single- and multi-scale models by describing them in a SBML-like markup language rather than by programming. Modelers interact with Morpheus through a user-friendly graphical interface (fig. 1) that facilitates the entire workflow from model construction to simulation, batch processing, visualization and analysis (Starruß et al., 2014). This enables also users with limited programming experience to construct and simulate models of multicellular systems. Its usability manages modeling workflows, enables rapid model development and makes Morpheus well-suited for use in education. Since the recent release of Morpheus 2.0 as an open-source framework, programmers can customize and extend the Morpheus framework. A flexible plugin interface allows a wide range of features to be added including new rules for cell shapes, behaviors and interactions as well as tools for visualization, analysis or data export. With these distinct interfaces, Morpheus separates modeling from programming, providing both usability for the modeler and flexibility for the programmer to create innovative models of multicellular systems. Source code and binary packages of Morpheus for Linux, Windows and Mac OSX are available from: http://imc.zih.tu-dresden.de/wiki/morpheus. Reference: J. Starruß, W. de Back, L. Brusch and A. Deutsch. Morpheus: a user-friendly modeling environment for multi-scale and multicellular systems biology. Bioinformatics, 30(9):1331-1332, 2014. Morpheus 2.0 Simulation browser Modeling editor PP 5-04 Modelling And Simulation Of Tumour Growth Compared With Xenograft Models And DbscTRAIL Therapy In 2D Cell Culture Population. Galliani, S.1, Shchekinova, E.1, Divine, M. 2, Reuss, M.1 University of Stuttgart, SRCSB - Stuttgart Research Center Systems Biology, Stuttgart, Germany Eberhard Karls Univeristy of Tübingen, Werner Siemens Imaging Center - Department of Preclinical Imaging and Radiopharmacy, Tübingen, Germany 1 2 High metabolic activity is the main trait that the majority of the tumours share. Functional imaging techniques, as Positron Emission Tomography (PET), are used in clinical oncology to identify tumours and metastasis in the body. A radiolabelled tracer as Fluoro-Deoxy-Glucose - FDG, injected i.v. in the patient, identifies areas of high glucose uptake, thus translating with PET scans the higher metabolic activity of tumour areas. Combining PET-FDG with MRI-ADC (Magnetic resonance imaging - Apparent Diffusion Coefficient) images, it is possible to localize inside the tumour areas of necrosis and proliferation. Once the tumour has been identified, the treatment with drugs or proper therapy are administered to the patient. We built a multiscale agent based model of 2D vascular tumour growth based on Perfahl et al. 2011 to investigate the role played by glycolysis (Pasteur and Warburg effect) on tumour development and to analyse the impact of the quiescent population. We then qualitatively compare the glucose uptake in our simulations results with the FDG-PET xenograft model images provided by Mathew Divine (Department of Preclinical Imaging and Radiopharmacy - Eberhard Karls University of Tübingen) identifying areas of proliferation and necrosis in the tumour. We accounted for necrotic areas formation with an age-dependent vessel collapse. Finally, given the experimental results of the effect of Db-scTRAIL molecule on cancer cell 89 Program / Abstracts cultures, we built a dose dependent function to model the apoptotic process in cells intercalated with the drug, together with randomly assigning, from a uniform distribution, the drug resistance to cells. We show that the shortage of glucose plays an important role in tumour development and subsequent angiogenesis. As soon as the tumour grows, the pressure exerted by the growing tumour damages the vessels infiltrated in the tumour tissue, thus leading to the vessels collapse and necrotic areas formation. Once applied the treatment with Db-scTRAIL, the initial population of cells develop a drug resistant ability. The subsequent population born after cell division show different resistance, thus effecting the overall growth/death dynamics. We observe that necrosis takes place in areas with insufficient delivery of nutrients leading to subsequent formation of stratified layers of necrotic, quiescent and proliferative tissues. The high heterogeneity of shape and size in mice’s tumour, despite their identical genetic background, show us the stochasticity of the tumour formation. A clear developing behaviour cannot be identified. The effect of Db-scTRAIL on 2D population of cancer cells is a function of dose-dependent apoptosis and drug resistance. References: Perfahl H., Byrne H.M., Chen T., Estrella V., Alarcón T., et al. Multiscale Modelling of Vascular Tumour Growth in 3D: The Roles of Domain Size and Boundary Conditions PLoS ONE, 6(4): e14790 (2011). PP 5-05 Data Needs Structure: Data and Model Management for Distributed Research Networks in Systems Biology and Systems Medicine Golebiewski, M.1, Krebs, O.1, Nguyen, Q.1, Owen, S. 2, Stanford, N. 2,3, Van Niekerk, D.4, Wolstencroft, K. 5, Bacall, F. 2, Kania, R.1, Rey, M.1, Weidemann, A.1, Wittig, U.1, Snoep, J. 2,3,4, Mueller, W.1, Goble, C. 2 HITS gGmbH, Heidelberg, Germany University of Manchester, School of Computer Science, Manchester, Great Britain 3 University of Manchester, Manchester Institute for Biotechnology, Manchester, South Africa 4 University of Stellenbosch, Department of Biochemistry, Stellenbosch, South Africa 5 Leiden University, Leiden Institute of Advanced Computer Science, Leiden, Netherlands 1 2 Systems Biologists need a data management infrastructure that enables collaborating researchers to share and exchange information and data as and when it is produced, throughout the entire iterative cycle of experimentation and modelling. We develop and offer integrated data management support for research in the fields of systems biology and systems medicine within and across research consortia. This comprises a whole package of solutions and is applied to geographically dispersed, interdisciplinary and large-scale research initiatives in which we are responsible for the scientific data management, like the German systems biology network Virtual Liver (http://www.virtual-liver.de/), from 2010 to 2015, and the newly established German research initiative ‘Systems Medicine of the Liver’ (LiSyM), as well as some European research networks like ERASysAPP (ERA-Net for Systems Biology Applications), the former SysMO initiative (Systems Biology of Microorganisms) or NMTrypI (New Medicines for Trypanosomatidic Infections). Parts of these solutions are also applied to smaller projects with a more local focus as the German project ‘Systems Biology of Erythropoietin’ (SBEpo) or the Synthetic Biology Centres at Manchester (SynBioChem) and Edinburgh (SynthSys). Our data management concept aims at bundling, storing and integrating research assets like data, models and description of processes and biological samples in a Findable, Accessible, Interoperable and Reusable (FAIR) manner (http://fair-dom. org). It consists of 4 major pillars: 1) Infrastructure backbone: The SEEK platform as registry and a commons for data, models, samples, processes and resulting publications or presentations, at the same time yellow pages for projects, people and events 2) Terminology: Tailored use of controlled vocabularies and ontologies to describe the data and its metadata (data describing the data) 3) Modelling support:Seamless handling and simulation of models by integrated modelling platforms (JWS-Online, SYCAMORE, Cytoscape) 4) Social support:Facilitators within the projects for gathering requirements and dissemination This concept is applied in our own research networks, but also used by other systems biology consortia. Unlike the majority of data management systems, we specifically support the interaction between modelling and experimentation. Datasets can be associated with models and/or workflows or biological samples, and model simulations can be compared with experimental data. 90 PP 5-06 An integrated temporal molecular response of vascular endothelial cells exposed to ionizing radiation Guipaud, O.1, Heinonen, M. 2, Buard, V.1, Tarlet, G.1, Jaillet, C.1, Chan, P. 3, Vaudry, D. 3, Vinh, J.4, d’Alché-Buc, F. 5, Milliat, F.1 IRSN, Human health radiation protection unit, Fontenay-Aux-Roses, France Aalto University, Department of computer science , Espoo, Finland 3 Rouen University, PISSARO Proteomic Platform, Rouen, France 4 ESPCI ParisTech, Spectrométrie de Masse Biologique et Protéomique, Paris, France 5 Paris-Saclay University, Télécom ParisTech, Paris, France 1 2 Question: Used in more than half of patients with tumors, radiation therapy (RT) plays a crucial role in the cure of cancers. The therapeutic index of RT depends on both tumor control and normal tissue tolerance. The vasculature participates in tumor progression and is required for normal tissue homeostasis orchestrating wound healing after radiation injury. We and others postulate that the vascular endothelium is a key compartment involved in the response of both normal and tumor tissues after RT and is the best clinical target to improve the differential effect of RT in the future. However, the response of vascular endothelial cell (EC) to radiation is not fully known and we lack an integrated view including consequences for both tumor and normal tissues. Here, we propose to use a systems biology approach to understand the dynamic response of ECs after irradiation and to decipher temporal biological networks involved in this response. Methods: We conducted exploratory experiments on primary human umbilical vein endothelial cells (HUVECs) under control, under a radiotherapy dose fraction (2 Gy) and under a high single dose (20 Gy). We measured proteomic profiles with 2D-DIGE and iTRAQ quantitative mass spectrometry, and transcriptional profiles of about 500 genes with real time quantitative PCR with measurements at 12 h, 1, 2, 3, 4, 7, 14 and 21 days following radiation exposure. We used the novel Bayesian likelihood ratio test we previously developed [1] to estimate the differential expression time periods, and we performed gene ontology and functional pathway analyses of the differential genes and proteins. Results: We applied the novel non-stationary Gaussian process (GPR, Gaussian Process Regression) as the underlying expression model, with major improvements on model fitness on perturbation and stress experiments. The method is robust to uneven or sparse measurements along time. Applying the ratio test to systems of genes and proteins provides the temporal response timings and durations of expression following irradiation. As a result, using the cascade of differential expression periods, domain literature and gene enrichment analysis, we gain insights into the dynamic response of endothelial cells to irradiation. Conclusions: The method estimates a temporal cascade of differentially expressed genes and proteins providing a largescale view on the expression progression of the irradiation response. Using statistical learning tools to unravel parts of the dynamics of biological networks involved in ECs in response to ionizing radiation, our global approach will tentatively help to bring together the problems of tumor destruction and protection of healthy tissues. [1] Heinonen, M. et al. Detecting time periods of differential gene expression using Gaussian processes: an application to endothelial cells exposed to radiotherapy dose fraction. Bioinformatics 31, 728-35 (2015). PP 5-07 Systems analysis of the structural and molecular changes along the dynamics of liver fibrosis development Hammad, S.1, Telfah, A. 2, Othman, A. 2, Hengstler, J. 3, Hergenröder, R. 2, Dooley, S.1 Medical Faculty Mannheim, Heidelberg University, Molecular Hepatology - Alcohol Associated Diseases, Department of Medicine II, Mannheim, Germany Dortmund University, Germany, ISAS-Institute for Analytical Sciences, Dortmund, Germany 3 Dortmund University, Germany, Leibniz Institute for Work Physiology and Human Factors (IfADo), Dortmund, Germany 1 2 Chronic liver disease (CLD) is a major health problem and it often leads to hepatocellular carcinoma (HCC). CLD progression is associated with altered tissue microarchitecture, gene expression and metabolite secretion. To study the progression of CLD, we used carbon tetrachloride (CCl4)-induced fibrosis and non-alcoholic steatohepatitis (NASH) -HCC model systems in mice. Adult male C57Bl6 mice were treated with CCl4 for 10 weeks (1 ml/kg body weight mixed with olive oil to 50 μl; twice per week i.p.). Then, 3-4 mice were sacrificed for plasma and liver analysis every week. For NASH-HCC model, on the second day after birth, male C57Bl6 mice were subjected to a single subcutaneous injection of 200 µg streptozotocin. Four weeks after injection, mice were fed high fat diet ad libitum until sacrifice at 6, 8, 12 and 20 weeks. At predefined time-points, blood and liver samples were collected for analysis. We have investigated the fingerprints of CLD progression by biochemical, histopathological and metabolites as well as gene expression analysis in a time-resolved experiment in both models. Biochemical parameters e.g. ALT and AST were increased during CLD progression in CCl4 model. This alteration was correlated with accumulation of extracellular matrix (ECM) as indicated by picro-sirius red and hematoxylin and eosin staining. Deposition of collagen, collagen-producing cells and lipid droplets in case of NASH-HCC are well correlated with CLD progression. Gene expression and metabolite analysis as well as three-dimensional reconstruction of liver microarchitectures are ongoing. Here we could show that our approaches to study liver systems medicine using CCl4 and NASH-HCC models developed CLD comparable to the human situation. 91 Program / Abstracts PP 5-08 Multiscale modeling of liver regeneration Hoehme, S.1, D’Alessandro, L. A. 2, Hengstler, J. G. 3, Klingmueller, U. 2, Drasdo, D.4,5 University of Leipzig, Institute of Computer Science, Leipzig, Germany German Cancer Research Center, Heidelberg, Germany 3 IfADo, Dortmund, Germany 4 INRIA Paris, Paris, France 5 University of Leipzig, IZBI, Leipzig, France 1 2 During the last years, modeling of different physiological and pathological aspects of the liver advanced significantly with the development of increasingly realistic models on molecular, cellular, tissue and whole organ scale. Nevertheless, model driven liver research is still hampered by a lack of techniques that allow robust integration of these different scales into unifying frameworks. We here present a novel multiscale spatio-temporal 3D model of liver tissue that is based on in-vivo 3D imaging and that may serve as such unifying framework. We use this model to study liver regeneration upon damage or tissue loss which depends on intracellular and tissue scale processes, which interplay with tissue mechanics. In order to capture all these processes at their respective scales, the presented multiscale modelling framework integrates sub-models at all relevant scales from intracellular signalling to body level. It thereby allows predictions on a wide range of possible regeneration scenarios and helps to identify on one hand particularly informative experiments permitting to distinguish between alternative mechanisms, on the other hand impossible scenarios that should not be pursued experimentally. In this way, the model predictions can guide the experimental strategy. The multiscale tissue model is able to simultaneously reproduce all experimental observations including the regeneration process kinetics on the tissue scale. The presented study is an example for how the tight systems-biological integration of experimentation and modelling, both covering multiple scales, can facilitate understanding of complex multi-scale processes as liver regeneration. PP 5-09 Quantification of the effect of mutations using a global probability model of natural sequence variation Hopf, T.1,2, Ingraham, J.1, Poelwijk, F. 3, Springer, M.1, Sander, C.1, Marks, D.1 Harvard Medical School, Systems Biology, Boston, United States TU München, Informatics, Garching, Germany 3 UT Southwestern Medical Center, Green Center for Systems Biology, Dallas, United States 1 2 Modern biomedicine is challenged to predict the effects of genetic variation. Systematic functional assays of point mutants of proteins have provided valuable empirical information, but vast regions of sequence space remain unexplored. Fortunately, the mutation-selection process of natural evolution has recorded rich information in the diversity of natural protein sequences. Here, building on probabilistic models for correlated amino-acid substitutions that have been successfully applied to determine the three-dimensional structures of proteins, we present a statistical approach for quantifying the contribution of residues and their interactions to protein function, using a statistical energy, the evolutionary Hamiltonian. We find that these probability models predict the experimental effects of mutations with reasonable accuracy for a number of proteins, especially where the selective pressure is similar to the evolutionary pressure on the protein, such as antibiotics. PP 5-10 Modeling Tumor-Immune System Interaction Jagiella, N.1, Schmidt, G.1 1 Definiens AG, München, Germany Question: New cancer immunotherapies aim on (re)enabling the bodies immune system to fight the cancer. As many of those therapies are expensive and linked to side-effects it is especially important to identify only those patients which they might be successful for. At the example of immune suppressor protein PDL1 we study the interactions between tumor and immune cells by a data-driven modeling approach with the goal of discriminating patient types by reinterpreting the data by parameter inference. Methods: PDL1 is expressed by immune cells to slow down the immune response, but can also be found on certain tumor cells allowing them to escape the immune system. As such it is target of anti-PDL1 immunotherapies. In a first step, we compare a compartment model to immunohistochemistry (IHC) and gene expression (GE) data from patients. The model describes CD8+ T cells and PDL1 expression in tumor cells by a system of moment equations derived from a master equation assuming the cellular processes to be stochastic. Then, by Bayesian methods we infer the model parameters from patient population snapshot data. 92 In a second step, we study the effects of the spatial arrangement of cell types using a hybrid multiscale model representing cells by an agent-based model and molecules by partial differential equations. Results: We found that the PDL1 expression rates inferred from the patient data helped to stratify the patients in respect to clinical outcome. Crucial was the assumption that PDL1 expression in tumor cells depends on IFNg secreted by CD8+ T cells, rather than direct contact with CD8+ T cells where no such effects could be seen. The spatial model was able to predict the ratio of CD8+ to PDL1+ cell densities found in the patient data by only one control parameter determining the patientspecific situation: the CD8+ T cell influx into the local tumor environment. Conclusions: From snapshot data of different sources we were able to identify the mechanisms controlling tumor-immune system interactions, specifically between CD8+ and PDL1+ cells. The interplay between short (PDL1) and long-range (IFNg) signaling was found to play a key role and might permit to identify potential responders to anti-PDL1 therapy. PP 5-11 Can in vivo 31P MRS assay of myocardial PCr/ATP ratio homeostasis test model predictions? Bakermans, O.1, Bazil, J. 2,3, Nederveen, A.1, Strijkers, G.1, Boekholdt, M.1, Beard, D. 3, Jeneson, J.4,1 Academic Medical Center University of Amsterdam, Radiology, Amsterdam, Netherlands Michigan State University, Physiology, East-Lansing, United States 3 University of Michigan, Molecular and Integrative Physiology, Ann-Arbor, United States 4 University Medical Center Groningen, Neuroimaging Center, Groningen, Netherlands 1 2 B PCr/ATP /ATP ratio myocardial PCr Main Objective: To integrate experimental Magnetic Resonance Spectroscopy (MRS) and numerical simulation platforms for the clinical investigation of human myocardial ATP free energy homeostasis. Here, we investigated if current standards of accuracy of in vivo 31P MRS assay of myocardial phosphocreatine (PCr) and ATP levels are adequate to test computational model predictions of myocardial PCr/ATP ratio homeostasis during metabolic stress in healthy versus diseased phenotypes. Materials and Methods: In vivo 31P Magnetic Resonance Spectroscopy (31P MRS) at 3 Tesla was used to measure the steadystate concentration ratio of PCr and ATP in the myocardium of 7 healthy subjects in resting state (heart rate ~60 bpm) using a linear 31P surface coil in combination with standard cardiac-triggered localization sequences1. Numerical simulations of myocardial PCr/ATP ratio homeostasis in response to metabolic stress resulting from increased cardiac work were performed using a biophysical model of myocardial oxidative metabolism 2. The model features a closed-loop homeostatic control design of mitochondrial oxidative ATP synthesis2. Altered myocardial ATP metabolism in sarcomeric cardiomyopathy (SCM) was modeled by a reduction in energetic efficiency of contraction. A Monte Carlo approach was applied to compute model output uncertainty using the empirical accuracy of the in vivo 31P MRS assay. Results: Using a 3D ISIS volume localization scheme1 to measure the in vivo basal myocardial PCr/ATP ratio in the healthy human heart (Figure 1A and B), a mean value of 1.33 ± 0.14 (± SD) was found. Alternatively, using 2D ISIS slice selection in combination with 1D SI read-out1, the myocardial PCr/ATP ratio was 1.97 ± 0.38 (mean ± SD). Bland-Altman analysis of repeatability yielded values of 42% and 44%, respectively. Numerical simulation showed that this level of accuracy is insufficient to test the validity of the metabolic network model and its central hypothesis that a closed-loop control design underlies myocardial ATP free energy homeostasis in the human heart (Figure 1C). Furthermore, detection of failing metabolic homeostasis in SCM may only be achieved if the cardiac work rate in patients is increased to levels associated with heart rates above 150 bpm (Figure 1C). Conclusions: A significant improvement in accuracy of in vivo 31P MRS assay of the myocardial PCr/ATP ratio in the human heart is needed towards validation of a computational model of myocardial oxidative metabolism and diagnosis of failing ATP free energy homeostasis in cardiomyopathy. This will require upgrading the MRS hardware (i.e., to a multi-element quadrature cardiac 31P coil and/or a 7 Tesla human MR scanner). References C 1. Lamb HJ et al. NMR Biomed 9: 217-27, 1996. 2. Wu F et al. J Physiol 586, 4193-4208, 2008. healthy SCM 93 Program / Abstracts PP 5-12 Multiscale modelling of hepatocellular carcinoma and transasrterial chemoembolisation therapy Jeneson, T.1, Lapin, A.1, Horger, M. 2, Bitzer, M. 3, Reuss, M.1 Stuttgart Research Center Systems Biology, Stuttgart, Germany University of Tuebingen, Diagnostic Radiology, Tuebingen, Germany 3 University of Tuebingen, Internal Medicine I, Tuebingen, Germany 1 2 Aim of study: To develop a multiscale model to mirror the development of advanced hepatocellular carcinoma (HCC) under transarterial chemoembolisation (TACE) therapy, propose improvements to the current therapy and predict their outcomes. TACE therapy: HCCs are characterised by high arterial blood flow unlike a healthy liver which receives majority of its blood from the portal vein [1]. In TACE therapy, drug eluting beads (DEBs) are injected through a catheter selectively into tumour vessels. The beads occlude these vessels thereby obstructing blood flow to the tumour and slowly release the drug into the tissue in a controlled manner. Selective injection of the beads and obstruction of vessels reduces washout of the drug from the tumour and damage to the surrounding healthy liver tissue. Modelling framework: A hybrid cellular automaton spanning the intracellular, cellular and tissue scales is used to model the development of vascular tumours [2]. In the model, cells are modelled as discrete quantities that move, proliferate and die. The cell cycle is governed by an ordinary differential equation. Predefined rules decide how cells move and die. Additionally, diffusible transport of nutrients like oxygen and glucose and growth factors like vascular endothelial growth factor (VEGF) are modelled using partial differential equations. Blood vessel network forms the vascular layer of the model. The vessels undergo structural adaptation at each time-step based on factors like the radius, flow, haematocrit, intra-vascular pressure etc. Results: We have extended the model to account for the dual blood supply (from the artery and the portal vein) as seen in a human liver. As expected, we observe, in our simulations, high arterial flow in the presence of a tumour. The location of DEBs is determined depending on the vessel radius, type of blood flow (arterial or venous) and the microenvironment. The blood flow in vessels with DEBs is obstructed and drug release is modelled as a slow release and diffusion process. Low blood flow leads to hypoxia and eventual necrosis due to collapse of blood vessels and apoptosis of quiescent tumour cells. Post therapy, we observe either a total eradication of the tumour or a revascularisation and relapse due to surviving tumour cells. Conclusions: We have been successful in modelling the development of HCC and the administration of TACE therapy. Model results showing either a complete response with no tumour cells post therapy or a partial response with some surviving tumour cells that lead to relapse of the tumour are consistent as seen in the clinical studies. Our future work aims at determining possible improvements to the current therapy, like administration of anti-angiogenic drug to prevent revascularisation post TACE and target the surviving tumour cells to avoid relapse or use of hyperbaric chambers to overcome hypoxia and increase the efficacy of the drug released from DEBs. [1] Breedis et al. The blood supply of neoplasms in the liver. The American Journal of Pathology, 30(5):969, 1954. [2] Perfahl et al. Multiscale modelling of vascular tumour growth in 3D: the roles of domain size and boundary conditions. PloS One, 6(4):e14790, 2011. 94 PP 5-13 Personalized liver function tests: A Multiscale Computational Model Predicts Individual Human Liver Function From Single-Cell Metabolism König, M.1,2, Marchesini, G. 3, Holzhütter, H. 2 Humboldt University, Institute for Theoretical Biology, Berlin, Germany Charité Berlin, Institute of Biochemistry, Berlin, Germany 3 University of Bologna, Department of Internal Medicine, Bologna, Italy 1 2 Understanding how liver function arises from the complex interaction of morphology, perfusion, and metabolism from single cells up to the entire organ requires systems-levels computational approaches. We report a multiscale mathematical model of the Human liver comprising the scales from single hepatocytes, over representation of ultra-structure and micro-circulation in the hepatic tissue, up to the entire organ integrated with perfusion. The model was validated against data on multiple spatial and temporal scales. Herein we describe the model construction and application to hepatic galactose metabolism demonstrating its utility via i) the personalization of liver function tests based on galactose elimination capacity (GEC), ii) the explanation of changes in liver function with aging, and iii) the prediction of population variability in liver function based on variability in liver volume and perfusion. We conclude that physiology- and morphology-based multiscale models can improve the evaluation of individual liver function. Personalized prediction of liver function (GEC) was implemented in a proof-of-principle app available at https://www.livermetabolism.com/gec_app/ A) Overview of detailed kinetic model of hepatic galactose metabolism in SBGN. B) Tissue-scale model of the sinusoidal unit comprising diffusion and convection based transport of substances in the sinusoid, diffusion-based transport of substances in the space of Disse and description of cellular metabolism via kinetic models of individual hepatocytes. Blood coming from the hepatic artery and portal vein enters the sinusoidal unit periportal and leaves pericentral. Transport between the sinusoid and the space of Disse occurs via fenestrations in the endothelial cells. Parameters and references are provided in thesupplement. C) Region of interests of the liver are modeled via the integration of multiple sinusoidal units based on the observed heterogeneity of structural parameters and microcirculation within the lobulus. D) Based on anthropomorphic information of subjects like age, gender, bodyweight and height the region of interests are scaled to the observed distributions of liver blood flow and liver volume. Reference values of galactose clearance (GEC) are calculated and the experimental value of GEC can be evaluated in this reference context. Based on available data on the distribution of anthropomorphic features the population variability can be evaluated. PP 5-14 Model-based predictions of inflammatory patterns in the breast lobular epithelium in relation to epithelial cell turnover Lopez Alfonso, J. C.1, Schaadt, N. 2, Wemmert, C. 3, Schönmeyer, Brieu, R.4, N.4, Forestier, G. 5, Feuerhake, F. 2, Hatzikirou, H.1 Center for Information Services and High Performance Computing (ZIH), Technische Universität Dresden, Innovative Methods of Computing, Dresden, Germany Institute of Pathology, Hannover Medical School, Hannover, Germany 3 Engineering science, computer science and imaging laboratory (ICube), Université de Strasbourg, Strasbourg, France 4 Definiens AG, Munich, Germany 5 Modelling, Intelligence, Process and Systems (MIPS), Université de Haute Alsace, Mulhouse, France 1 2 Question: Most of the factors and mechanisms that regulate the inflammatory microenvironment in breast lobular epithelium are only partly understood. For instance, those related to lymphocytic lobulitis (LLO), a characteristic pattern of inflammation characterized by lymphoid cells infiltrating lobular structures. Interestingly, abnormal inflammatory patterns are not only detected adjacent to clinically manifest hereditary breast cancer, but can also be observed in prophylactically removed non-neoplastic breast tissues from patients without or with familial genomic aberrations associated with a high 95 Program / Abstracts cancer risk. This led to the hypothesis that LLO could be associated with genomic aberrations that impair DNA repair. Thus, understanding the biological and functional role of inflammation is critical to evaluate whether alterations in the breast function could associate with the development of cancer. Methods: To gain new insights into the role of inflammation in the breast epithelium, we developed a dynamical multiscale agent-based model for the interactions between immune and epithelial cells in lobular epithelium during the menstrual cycle. Physiological model parameters were calibrated on the basis of spatial data extracted from digital whole-slide images of immunohistochemical epithelial, vascular and immune cell markers. These images were acquired on clinical annotations from a cohort of healthy patients who underwent reduction mammoplasty. To better align the mathematical model with microscopic observations, we developed a modular workflow combining a convolutional neural network to detect regions of interest with an auto-adaptive random forest pixelwise classifier to detect nuclei (“nucleus container” module; Definiens, Germany). This immune cell analysis provided the quantitative read-out required as input data of the proposed model. Results: We found that the immunological context as defined by the density, functional orientation and spatial distribution of immune cells, provides valuable information to predict potential pathological scenarios. In addition, while epithelial damage due to DNA repair defects was predicted to induce clustering patterns of immune cells, high hormone levels result in spatial distributions where clusters were not present. The length of the menstrual cycle was found to be a crucial patientspecific factor shaping inflammatory responses to malignant changes in breast tissue. Conclusions: Modeling results evidence that spatial correlations between infiltrating lymphocytes and personalized clinical data may contribute to the extension of criteria for biopsy evaluation, supporting the development of novel prognostic markers and opening new perspectives for immunomodulatory therapeutic interventions. PP 5-15 Modeling disease Progression in Myeloproliferative Neoplasms, a Systems Medicine Approach Montazeri, M.1, Brehme, M.1, Schuppert, A.1, Koschmieder, S. 2, Brümmendorf, T. 2 1 2 Joint Research Center for Computational Biomedicine, Aachen, Germany Department of Hematology, Oncology, Hemostaseology, and Stem Cell Transplantation, faculty of Medicine, RWTH Aachen, Aachen, Germany Myeloproliferative neoplasms (MPN) are molecularly well-defined stem cell disorders, which can be studied to investigate multi-step malignant transformation. Different MPN subtypes such as polycythemia vera (PV), essential thrombocytosis (ET), and primary myelofibrosis (PMF) are characterized by eventual disease progression from a “benign” chronic state towards myelofibrosis and leukemic transformation. The identification of biomarkers predictive of the individual disease progression risk is crucial for therapy success, but challenged by the heterogeneity of the cell populations involved, the complexity of the underlying mechanisms and their mutual interactions. Our recent study on disease progression modelling in CML revealed that combining two patient-derived genomic scores, CD34+ similarity and gene expression entropy, creates a direct link between patients and mechanistic disease models. This link allowed estimating patients’ disease progression state with respect to disease evolutionary time as well as risk of transition from chronic to malignant phase. We are extending this concept to MPNs in order to develop predictive models for the onset of instability during early-stage disease, which leads to progression to more developed stages. To develop such prognostic biomarkers we are studying the differences between genetic and epigenetic properties in early vs. late MPNs, including the direct comparison to clinical parameters. Our goal is to translate these findings into a predictive mathematical model of disease progression. We will perform comparative omics analyses, including gene expression array, NGS, Chip-seq and genome-wide DNA methylation screens using clinical biopsies obtained from patients in the MPN-SAL registry. Preliminary analyses of publicly available data comparing patient whole blood vs. peripheral blood neutrophil gene expression by principle component analysis (PCA) resolved differential gene expression between ET and PV, as surrogates for “early” MPN, versus PMF as a “late” MPN. Comparing PMF vs. controls in whole blood, > 400 genes are separated from residual space of PCA compared to only ~150 in neutrophils. We hypothesize that these differences indicate putative population dynamic effects reminiscent of those observed during CML progression. In parallel we are investigating clinical parameters to find a set of variables capable of classifying different MPNs and time-wise prediction of transformation from chronic to acute stages. 96 PP 5-16 Development of a Multiscale Systems Biology Approach to Study Atherosclerotic Plaque Progression in WTLdlrKO and M-S196ALdlrKO mice: Integrating Mathematical and Murine Models Pichardo-Almarza, C.1,2, Gage, M. 3, Pineda-Torra, I. 3, Diaz-Zuccarini, V.1,2 UCL, Mechanical Engineering, London, Great Britain UCL, Institute of Biomedical Engineering, London, Great Britain 3 UCL, Centre for Clinical Pharmacology, London, Great Britain 1 2 Main objective of the study. Risk factors for atherosclerosis (ATH) have been extensively studied over the last two decades using murine models in order to understand disease mechanisms. The aim of this work is to develop a Systems Biology approach to understand atherosclerotic plaque (ATHP) progression in WT LDLR-KO and M-S196A LDLRKO mice. A multiscale modelling approach is proposed, where different biological scales are used to dynamically simulate and understand different aspect of the disease. Materials and Methods. Diet induced ATH: Mice were fed Breslow Western diet (21% fat from milk, supplemented with 0.15% wt/wt cholesterol) for 12 weeks from 8 weeks of age to induce ATH. Metabolic tests: Plasma total cholesterol was determined via colorimetric assay. Plasma HDL and LDL/VLDL levels were measured via quantitative colorimetric assay. Systemic inflammation: Plasma MCP-1 was determined by enzyme-linked immunoassay. ATH: Mice were perfusion-fixed with phosphate-buffered paraformaldehyde (4% [wt/vol.], pH 7.2) under terminal anaesthesia. The entire aortic tree was dissected free of fat and other tissue. Aortae were stained with oil red O and mounted on a glass slide before imaging under a dissection microscope. Aortic root (AR): Hearts were paraffin embedded and 5 μm AR sections were stained with H&E. Mathematical Modelling and Simulation: The multiscale mathematical model used is based on previous work proposed for humans [1] where temporal evolution of inflammatory processes and ATHP progression are modelled using differential equations. Parameter fitting was made via literature search and data provided by the experimental team. The model was implemented in Matlab®. Results. Preliminary results from the model match reasonably well the data reported from experiments: AR of M-S196A LDLRKO mice showed higher plaque coverage (around 14%) with respect to WT LDLR-KO mice (11% coverage) after 20 weeks. Two set of parameters were used to simulate different types of mice, allowing to introduce new variables in the model related to diet, lipid levels (e.g. LDL-cholesterol) and how the ATHP growth is finally affected by these variables. Conclusion. These first results are encouraging as this approach helps to quantitatively understand biological mechanisms in ATHP progression in WT LDLR-KO and M-S196A LDLRKO mice. The disease was presented as a dynamic process (i.e. evolving in time) and the model helped understanding how different physiological mechanisms and parameters are related in the organism as a whole. New in-silico experiments are being designed in order to help to design new experiments with the murine models. References [1] Pichardo-Almarza C, Metcalf L, Finkelstein A, Diaz-Zuccarini V. Using a Systems Pharmacology Approach to Study the Effect of Statins on the Early Stage of Atherosclerosis in Humans. CPT Pharmacomet Syst Pharmacol. 2014 Dec 30;4(1):4150. PP 5-17 3D multiscale modeling of vascularized tumor development inside colon Shchekinova, E.1, Galliani, S.1, Joshi, T.1, Reuss, M.1 1 Stuttgart Research Center Systems Biology, Stuttgart, Germany Main objectives: The aim of our study was to model the growth of tumour and vasculature within colon tissue using particular setup of vasculature and an initial configuration of the miscroscopic tissue that resembles the polyp structure within colonic cript. Methods: We used three-dimensional multiscale model [Perfahl2011,Galliani2015] of vascular tumour growth. To study the impact of choice of initial vasculature and density of tumour cells on the subsequent evolution of tumour within tissue we performed a set of stochastic simulations with variable intial density and uniform distibution of vessels within tissue. In particular, the angiosenesis and vasculature formation were compared for simulations using Gaussian density distribution of initial tumour cells within heathly tissue. Modeling of tumour development inside the colon tissue was performed using colon polyps associated with tubular and villous structure. Results: To get an idea about vasculature initiation we neglected epithelial dynamics but looked at the development of vascularized tumour on a larger scale. For this purpose a small tumour was implanted into tissue with developed vasculature network. The initial vasculature network was modelled by taking different initial three--dimensional geometry that resemble the colon polyp configuration. We investigated the sensitivity of vasculature development on chemotaxis coefficient for the edothelial sprout. The higher density of the vascular network in this case provided better conditions for the subsequent tumour spread after the tumour implantation. The tumour stimulates new sprouting and migration of endothelial cells inside 97 Program / Abstracts it. Initially the degradation of a healthy tissue was observed, at a later time the model predicted fast development of tumour with a final invasion into muscularis mucosa layer. Conclusion: Simulation results for two different initial setups showed similar behavior of tumour proliferation. Stochastic simulations for 3D slices of tissue showed that while proliferation rates of tumour cells were similar across different realizations, large variability was observed in propagation of the tumour at boundary layer. Reference Galliani et al., 2015. From preclinical biological models to human colon cancer: A contribution from mathematical modeling. Poster at the 2nd International Conference on Computing, Mathematics and Statistics (iCMS2015) doi: 10.13140/ RG.2.1.4415.3686 Perfahl et al, 2011. Multiscale Modelling of Vascular Tumour Growth in 3D: The Roles of Domain Size and Boundary Conditions. PloS ONE 6(4),e14790. doi:10.1371/journal.pone.0014790. PP 5-18 Interplay of nucleosome positioning, covalent modifications and transcription factor binding Teif, V.1 University of Essex, Colchester, Great Britain 1 Binding of transcription factors (TFs) to the genome is a central determinant of gene regulation. In higher eukaryotes, it is dependent on both the DNA sequence and the chromatin environment. On the other hand, differential binding of special TFs, such as a CCCTC-binding protein (CTCF), can dramatically affect 3D chromatin structure. Thus, two cells of the same organism, containing the same genome, can have very different TF binding depending on the cell state, which determines differences in their gene expression. In several ongoing projects we have narrowed down the problem of predicting celltype specific TF binding by focusing at a single protein CTCF. CTCF represents a prototypic case of a context-dependent DNA-interacting protein. While it can act as a tumour suppressor, its overexpression can also favour tumour progression in breast cancer. In many cases, CTCF exerts its function by demarcating boundaries between thousands of chromatin domains and establishing 3D chromatin structures through DNA loop formation. Large fraction of cell-type specific variability in CTCF binding to the several thousands of its potential binding sites in the human genome is linked to changes in DNA methylation and nucleosome repositioning. However, in most cases differential DNA methylation is not the cause, but rather a consequence of the differences in CTCF binding, while the cause of CTCF binding differences in some cases is linked to covalent modifications of CTCF itself. I will provide an update about our recent results and current understanding of this system. References Pavlaki I. , Docquier F., Chernukhin I., Teif V. B., Klenova E. (2016). Poly(ADP-ribosyl)ation dependent changes in CTCF DNA binding and gene expression patterns. In preparation Teif V.B., Beshnova D.A., Marth C., Vainshtein Y., Mallm J.-P., Höfer T., Rippe K. (2014). Nucleosome repositioning links DNA (de)methylation and differential CTCF binding during stem cell development. Genome Research 24, 1285-1295. Beshnova D.A., Cherstvy A.G. Vainshtein Y., Teif V.B. (2014). Regulation of the nucleosome repeat length in vivo by the DNA sequence, protein concentrations and long-range interactions. PLoS Comput. Biol. 10(7):e1003698. Teif V.B., Vainshtein Y., Caudron-Herger M., Mallm J.-P., Marth C., Höfer T., Rippe K. (2012) Genome-wide nucleosome positioning during embryonic stem cell development. Nature Struct. Mol. Biol. 19, 1185-92. PP 5-19 Modelling of tertiary lymphoid organ development in the context of kidney transplant. Uvarovskii, A.1, Schaadt, N. 2, Feuerhake, F. 2, Meyer-Hermann, M.1,3 Helmholtz Centre for Infection Research, Department of Systems Immunology and Braunschweig Integrated Centre of Systems Biology, Braunschweig, Germany Institute for Pathology, MHH, Hannover, Germany 3 Institute for Biochemistry, Biotechnology and Bioinformatics, TU Braunschweig, Braunschweig, Germany 1 2 Question: Tertiary lymphoid organs (TLO) are organized lymphoid structures which develop at locations other than lymphoid tissue. A matured TLO consists of adjacent T and B cell zones, follicular dendritic cell (FDC) network and may even include developed germinal centres. TLOs may be found at sites of chronic inflammation, for example, in transplanted organs. Mouse transplantation models showed that the presence of TLO in skin grafts increased the risk of rejection. Herein, we consider TLO development in the context of transplanted kidney, where we observed lymphocyte clusters adjacent to blood vessels in human biopsy samples and would like to understand dynamic implications of the snapshots seen in the biopsies for the long-term graft outcome and treatment decisions. Methods: Since key processes in the system are guided by molecule interactions, what includes diffusion and binding kinetics, they may be described by mathematical modelling. We developed spatio-temporal agent-based model which consists of several levels of abstraction: cell movement and collision detection, chemokine diffusion and receptor dynamics at the cell 98 surface. B cells are represented as spheric agents performing chemotactic migration towards sources of CXCL13, a chemokine secreted by activated stroma cells (FDCs). FDCs were shown to derive from perivascular precursors upon lymphotoxin receptor interaction, probably, with B lymphocytes. In our model stroma cells measure an hypothetic signal based on contacts with B cells and start to secrete CXCL13 as soon as they sense the threshold signal value. CXCL13 binds to receptors on B cells, which are being internalized, resulting in chemokine consumption and change in surface receptor number. Hence, CXCL13 distribution is described by the diffusion equation coupled to receptor dynamics. Results: We estimated receptor densities and secretion rates necessary for B cell attraction for cases of a single FDC and for an FDC cluster and what maximal lymphocyte cluster size can be generated by this systems. We show that for rates in the literature range, chemokine degradation does not influence its gradient at the distances of TLO size order. Conclusion: The simulations may be used in adjustment of anti-chemokine therapy for destabilizing cell clusters. Biopsy comparison may reveal critical B cell density for stroma activation and will help to estimate probability of outcomes. PP 5-20 Parallel analysis of the proteome, phosphoproteome and N-terminome to characterize altered platelet functions in the human Scott syndrome Solari, F.1, N. Mattheij, J. A. 2, Burkhart, J. M.1, Swieringa, F. 2, Sickmann, A.1, Heemskerk, J. W. M. 2, Zahedi, R.1 1 2 ISAS , Dortmund, Germany CARIM, Maastricht, Netherlands The Scott syndrome is a very rare and likely under-diagnosed bleeding disorder associated with mutations in the gene of anoctamin-6. Platelets from Scott patients are impaired in various Ca 2+-dependent responses including phosphatidylserine exposure, membrane ballooning and intracellular protein cleavage, resulting in an impaired procoagulant response. Owing to its central role in thrombosis and hemostasis, we aimed to improve our understanding of the molecular mechanisms of the procoagulant response by studying Scott platelets using quantitative proteomics. We hypothesized that alterations in protein expression, protein phosphorylation, and proteolytic cleavage may contribute to this complex phenotype. Thus, we applied a multipronged iTRAQ-based proteomics to quantify changes between healthy controls and Scott patients (i) in the platelet proteome, as well as upon platelet activation in (ii) the phosphoproteome and (iii) N-terminome to identify Calpain substrates. Only a limited number of proteins had decreased (70) or increased (64) expression in Scott platelets. Parallel reaction monitoring confirmed that anoctamin-6 was absent, while interestingly aquaporin-1 was highly upregulated. Quantification of 1,574 phosphopeptides revealed major differences between Scott and control platelets after stimulation with thrombin/ convulxin or ionomycin. In the patient platelets phosphorylation levels were increased in multiple proteins regulating the cytoskeleton or signaling events. Moreover, 1,596 terminal peptides were quantified in activated platelets, 180 of which were confirmed to be calpain-regulated, whereas a distinct set of 23 neo-N-termini was identified as caspase-regulated. Notably, in Scott platelets calpain-induced cleavage was down-regulated, including cytoskeleton-linked and signaling proteins, which is in accordance with the increased phosphorylation state. Thus, sensitive multi-pronged quantitative profiling of the proteome, phosphoproteome and N-terminome provided multilayer insights into aberrant signaling in Scott syndrome that can be related to the partial protection of Scott platelets to the detrimental Ca 2+-dependent cytoskeleton and membrane changes that are normally associated with phosphatidylserine exposure. 99 Program / Abstracts PP 6: Systems Medicine & Systems Pharmacology PP 6-01 SIMPLEX: a combinatorial multimolecular omics approach for systems biology Ahrends, R.1, Coman, C.1 1 ISAS, Lipidomics, Dortmund, Germany Question - Interconnected molecular networks are at the heart of signaling pathways that mediate adaptive plasticity of eukaryotic cells. To gain deeper insights into the underlying molecular mechanisms, a comprehensive and representative analysis demands a deep and parallel coverage of a broad spectrum of molecular species. Methods - Therefore, we introduce SIMPLEX (SImultaneous Metabolite, Protein, Lipid EXtraction procedure), a novel strategy for the quantitative investigation of lipids, metabolites and proteins. Compared to unimolecular workflows, SIMPLEX offers a fundamental turn in study design, since multiple molecular classes can be accessed in parallel from one sample with equal efficiency and reproducibility. Results - Application of this method in mass spectrometry based workflows allowed the simultaneous quantification of 360 lipids, 75 metabolites and 3327 proteins from 10 6 cells. The versatility of this method is shown in a model system for adipogenesis - PPARG signaling in mesenchymal stem cells - where we utilized SIMPLEX to explore cross-talk within and between all three molecular classes and identified novel potential molecular entry points for interventions, indicating that SIMPLEX provides a superior strategy compared to conventional workflows. Conclusions - The identification of a close collaboration between the PPARG signaling on the protein level and TAG metabolism on the metabolite level proves our initial motivation to investigate the system within the cellular context rather than a restricted single-sided way and allowed for the elucidation of a further level of feedback based PPARG control derived from the lipid level. We established a protocol that can deliver a representative picture of an entire biological system by extending the detection capabilities to more than one molecular class. As mass spectrometry instrumentation continuously advances regarding sensitivity and speed, we expect that this novel workflow will soon accommodate more quantitatively accessible molecular species and will deliver new exiting perspectives into metabolic disorders. PP 6-02 Baseline Matters: the Effect of Initial Immune State on Outcome of Septic Patients Alpert, A.1, Mansour, A. 2, Starosvetsky, E.1, T. Sweeney, E. 3, Menck,K. 2, Hinz, J. 2, Khatri, P. 3, Shen Orr, S.1 Technion, Haifa, Israel Georg-August-University, Anaesthetics, Goettingen, United States 3 Stanford, School of Medicine, Stanford, United States 1 2 Sepsis is a severe life-threatening systemic inflammatory response caused by infection that is responsible for more than 250,000 deaths annually in the US. The immune response in sepsis is composed of two opposing components, namely immune activation followed by immune suppression, which is dominated by T-cell exhaustion. In spite of wide variation in the immune system between individuals, the effect of the initial immune state on outcome in sepsis is still unknown. Using CyTOF, we monitored at high resolution the peripheral-blood dynamics of septic patients, including abundances of cell subsets and functional signaling responses to IL-6 stimulation. We identified major shifts in the T-cell compartment which occur gradually during sepsis, including an increased expression of inhibitory receptors and a decreased expression of co-stimulatory molecules, both hallmarks of T-cell exhaustion. Exhausted CD8+ T-cell subsets responded poorly to IL-6 stimulation, reflecting a functional defect of exhausted T-cells in sepsis, and suggesting baseline levels of exhausted cells may affect outcome. To test this, we computed a T-cell exhaustion enrichment signature at baseline hospitalization and over time in a large gene expression dataset of trauma patients that contracted nosocomial infections. We identified a significant correlation of severity and baseline exhaustion enrichment, measured less than one day following trauma. Our findings highlight the role of initial immune state on determination of clinical outcome and suggest that the baseline immune-state variability may be a useful diagnostic tool for identification of septic patients at risk. 100 PP 6-03 Dynamics of the tumor-infiltrating lymphocyte repertoire in melanoma and pancreatic cancer Appel, L. M.1,2, Poschke, I. 3, Faryna, M.4, Hassel, J. 5, Strobel, O.6, Diken, M.7, Kranz, L. M.7, Offringa, R. 3, Höfer, T.1,2, Floßdorf, M.1,2 German cancer research center (DKFZ), Theoretical Systems Biology, Heidelberg, Germany BioQuant Center, University of Heidelberg, Heidelberg, Germany 3 German cancer research center (DKFZ), Molecular Oncology of Gastrointestinal Tumors, Heidelberg, Germany 4 BioNTech Diagnostics, Mainz, Germany 5 Heidelberg University Hospital, Department of Dermatology, National Center for Tumor Diseases (NCT), Heidelberg, Germany 6 Heidelberg University Hospital, General, Visceral and Transplantation Surgery, Heidelberg, Germany 7 Institute for Translational Oncology (TrOn), Mainz, Germany 1 2 In recent years immunotherapeutic approaches with the goal to harness the patient’s T cell response have become the standard therapy for many cancers. Still, the unpredictability of the individual response to a particular immunotherapeutic approach calls for a better understanding of the underlying mechanisms informing the identification of biomarkers for patient stratification. To further our mechanistic insights, we are devising a mathematical model of the anti-tumor T cell response in the B16-OVA melanoma mouse model. Proliferation monitoring and accurate cell counting in all relevant compartments (tumor, blood, draining lymph nodes, spleen) in conjunction with the mathematical model allows us to quantify the time-dependent proliferative activity of the tumor-infiltrating lymphocytes and their migration behaviour. In parallel, we are investigating next generation sequencing data of the repertoire of tumor-infiltrating T cells in melanoma and ductal adenocarcinoma (PDA) patients. We have optimized the sequencing protocol for our purposes, such that it provides an accurate, quantitative representation of the clone sizes. For melanoma, immune therapy is often successful and we are using these data to identify potential biomarkers predicting clinical outcome. Preliminary results of our analysis of the TIL repertoire in PDA patients indicate similarities to that in melanoma, suggesting a hitherto unappreciated tumor-reactivity of the infiltrating lymphocytes; immunotherapeutic interventions like adoptive T-cell therapy might hence be a treatment option also in this type of cancer. PP 6-04 Markov-Chain Monte-Carlo methods to analyze mechanistic disease simulators: Applicability and shortcommings Ballnus, B.1, Müller, C. 2, Hatz, K. 3, Hug, S.1, Weyßer, F. 3, Theis, F.1, Schuppert, A. 2, Hasenauer, J.1, Görlitz, L. 3 Institute of Computational Biology, Helmholtz-Zentrum München, Neuherberg, Germany RWTH Aachen, AICES Graduate School, Aachen, Germany 3 Bayer Technology Services, Leverkusen, Germany 1 2 Main Objectives: Disentangling the complex interplay between development of pathophysiological and drug-induced changes on clinical phenotypes by mechanistic models is a powerful approach towards personalized therapies of complex diseases. For pharmacokinetics of drugs this paradigm was already successfully developed (e.g. [1]). Pharmacodynamic approaches still fail as mechanistic disease models typically contain complex dynamic features hindering applicability of developed approaches. In this case-study, we examine the applicability of Bayesian inference and Markov-Chain Monte-Carlo (MCMC) methods for two mechanistic models. The first model is a cardio-vascular whole body model (CV), which models human in vivo mechanisms of blood flow & pressure on different time scales (based on [2,3]). The second application is a blood coagulation model (BC), which aims to predict the coagulation behavior (based on [4]). Materials & Methods: Both models are described by large ordinary differential equation systems. Model parameters, their variability and uncertainty have to be inferred from experimental data. The experimental data contain measurements of physiological values in humans for CV and total Thrombin in human blood plasma mixtures for BC. Inference of the parameters and their distribution from the models is done using MCMC algorithms (e.g. [5]). Results: Model analysis indicates that both simulators possess strong non-linear behavior, bifurcations and periodic or a-periodic dynamics. These properties are shown to lead to instabilities of common MCMC approaches. We show that a preceding optimization leads to more robust estimation results. Best results were obtained for BC using classic MetropolisHastings (MH) in combination with convex priors and a locally estimated covariance. However, our posterior approximations revealed multiple, weakly separated modes. In addition, our CV related analysis revealed that periodic solutions can induce multi-modalities. Furthermore, our results show that bifurcations have less impact onto MCMC algorithms then initially expected. In analogy, the impact of multistabilities seems negligible. However, systems with separated, approximately equally weighted modes are only targetable by more sophisticated multi-chain approaches as Parallel Tempering or Parallel Hierarchical Sampling. Conclusions: We have confirmed the existence of practical limits using classic MCMC approaches in real-life applications. This prevents us from inferring parameter densities within finite time. More sophisticated MCMC methods or augmentations 101 Program / Abstracts based on previous optimization are necessary to reliably approach these problems. The latter improvement has been successfully applied in previous pharmacokinetic studies and our BC results without major additional computational costs. PP 6-05 Modeling biliary fluid dynamics reveals possible mechanism for dose-response and personalization of UDCA treatment in PSC Ostrenko, O.1, Segovia-Miranda, F. 2, Brosch, M. 3, Erhart, W.4, Meyer, K. 2, Kretzschmar, G. 5, Jüngst, C.6, Lammert, F.6, Zerial, M. 2, Schafmayer, C.4, Brusch, L.1, Hampe, J. 3 Technische Universität Dresden, Center for High Performance Computing (ZIH), Dresden, Germany Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany 3 Technische Universität Dresden, University Hospital, Dresden, Germany 4 University Hospital Schleswig Holstein, Kiel, Germany 5 Technische Universität Dresden, Molecular Cell Physiology and Endocrinology, Institute of Zoology, Dresden, Germany 6 Saarland University Medical Center, Homburg, Germany 1 2 Background and Question: Primary sclerosing cholangitis (PSC) is a progressive liver disease characterized by fibroobliterative destruction of the intra- and/or extra-hepatic bile ducts. Combined with immune-mediated insults to the bile duct, biliary flow obstruction might lead to pressure damage to the biliary epithelium and may drive further disease progression. Currently, the only and controversial medical treatment is the choleretic drug ursodeoxycholic acid (UDCA). However, while moderate doses (10-15mg/kg/day) of UDCA might improve liver function tests and histology, high-dose UDCA has been demonstrated to increase mortality in a RCT, thereby calling UDCA treatment in question. Objective: Develop a hydrodynamic model of biliary pressure and flow to assess the effect of UDCA on biliary pressure in normal liver and PSC. Methods: A recently developed theory of bile secretion and transport was applied to UDCA treatment. UDCA reduces bile viscosity (1) and increases solute concentration and consequently osmotic water influx into bile (2). Model parametrization was completed using measured viscosity from gallbladder bile of patients treated with UDCA and quantitative 3D-reconstruction of stacked high resolution microscopy of healthy human and PSC livers. Bile viscosity and osmotic water inflow are modelled as functions of UDCA dose according to effects (1) and (2) above. Results: This quantitative model allows the assessment of the mechanical consequences of UDCA treatment in a dosedependent manner and predicts intrahepatic fluid pressure as a function of the physical properties of bile and the geometrical properties of the tissue. For healthy liver, the model reproduces the insensitivity of biliary pressure to UDCA dose. In PSC, a pronounced dose-response curve is predicted by our model: biliary pressure is decreasing at UDCA dose increments, reaches a minimum and then rapidly increases beyond baseline upon further dose increments. The optimal UDCA dose corresponding to the maximum pressure reduction can be quantified as a function of the physical and geometrical parameters that may vary from patient to patient. Conclusions: This study provides a hydrodynamic explanation of the clinical dose-response of UDCA in PSC. Further, it confirms tolerance to high doses of the compound in normal liver. Mechanistically, the UDCA-induced increase in water inflow has pressure-lowering (bile dilution) and pressure-increasing (increased fluid volume in need of drainage) consequences. Thus, the model advocates individualized dose-optimization based on the patient-specific biliary microgeometry, thereby potentially rendering UDCA treatment safer and more widely applicable. PP 6-06 Improving non-invasive liver function diagnostics Bulik, S.1 Charité - Universtätsmedizin Berlin, Institut f. Biochemie / Computational Biochemistry, Berlin, Germany 1 The liver is the central organ of detoxification and metabolic homeostasis while at the same time exposed to pathogenic and toxic challenges due to its position in the body. A peculiarity of liver diseases is their often silent progression without serious clinical symptoms until they reach a threshold stage characterized by a sudden and often life-threatening onset of severe liver failure. Improving liver diagnostics will help to detect the developing decline of liver function and enable the application of required therapies. We propose an adaptation of existing CO2 based breath-tests to individual bicarbonate metabolism to improve the sensitivity of the tests. Substances containing stable isotope labeled 13C components like methacetin or galactose are applied in clinical routine to assess the liver function by the measurement of 13CO2 expiration. The impact of the individual CO2 distribution in the body is hitherto neglected. We developed the 2DOB test where this effect is separately measured and by computational modelling the functional liver test is purged from this influence. Furthermore, the combination of functional tests with structural data from magnetic resonance elastography (MRE) will 102 provide a tool to develop a spatially resolved 3D-liver function model generating a means to support clinical assessment of liver disease severity and control of disease progression. PP 6-07 Integrative analysis of chemical high-throughput screens uncovers novel biological information. Liu, X.1, Adornetto, G.1, Campillos, M.1 1 Helmholtz Zentrum Munich, Institute of Bioinformatics and System Biology, 1, Germany Objectives of the study: Chemical high-throughput screens (HTS) have been extensively used by pharmaceutical companies, as well as by academic research groups to identify compounds with activity on protein targets or with specific phenotypic activity. Recently, an extensive number of these screens have been deposited on public databases such as PubChem and ChemBank. Concurrently, large databases of drug targets have appeared in the public domain. This explosion of chemical information in the public sector is enabling the integration of information from HTS and more important, its systematic exploration. We have developed computational methods for the integration of chemical HTS, applied them to the ChemBank repository and interrogate the resulting resource. In this work, we illustrate the capabilities of this integrative resource to uncover novel biological information with an example showing that chemical screens sharing selective hits uncover novel biological relationships. Methods: To enable the integration of disparate chemical HTS, we have improved existing computational methods for the systematic identification of hits in chemical screens, developed the target prediction tool HitPick, and created filters to remove promiscuous compounds. In order to compare the biological activities measured in the screen, we have assigned Gene Ontology categories to the screens and constructed chemical hit profiles with sets of compounds differing on their selectivity level for 1640 screens of ChemBank repository. Results: We hypothesized that if the same set of active compounds modulates two apparently unrelated biological processes, these processes are likely to be connected through the common targets of the compounds and thus, may be unexpectedly related. We have tested and proved the hypothesis using the set of 1640 chemical genetic assays stored in ChemBank. We showed that assays that share hits also share biological associations through common protein targets. We suggested novel biological activities of known drug targets that represent potential targets for the treatment of cancers such as the Cannabinoid receptor 2. Conclusion: With this study we demonstrated the potential of integrative computational approaches applied to chemical HTS to uncover novel biological information. PP 6-08 A NAT2 Pharmacogenomic based PBPK model of Isoniazid in Men and Its Application in Adjusting Tuberculosis Chemotherapy Cordes, H.1, Thiel, C.1, Baier, V.1, Blank, L.1, Kuepfer, L.1 1 RWTH, iAMB, Aachen, Germany Due to its greatest early bacterial activity among fist-line anti-tuberculosis agents, isoniazid plays an essential role in tuberculosis treatment. Polymorphisms in the NAT2 gene encoding for the N-acetyl transferase 2 cause a trimodal distribution of isoniazid pharmacokinetics (PK) among human populations, categorized into slow, intermediate and fast acetylators. Success of isoniazid chemotherapy is hence associated with acetylator status and general health state of a patient. Still, the WHO recommended standard dose is administered regardless of acetylator type or immune status of patients. Adverse effects during isoniazid therapy occur in about 5 % - 33 % of all patients. Clinical studies show that slow acetylators have a higher risk to develop drug-induced toxicities, while lower treatment efficacies are found for fast acetylators and immune deficient patients. To mechanistically assess this trade-off, we developed a physiologically based pharmacokinetic (PBPK) model to describe the NAT2-dependent PK of isoniazid and five of its metabolites. The comprehensive PBPK model was combined with a pharmacodynamics (PD) model, describing the anti-bacterial drug effect on mycobacterial growth in the lung. The resulting PBPK/PD model allowed the simultaneous consideration of treatment efficacy on the one hand and exposure to toxic metabolites in off-target organs causing drug induced side effects on the other. Aim of the study hence was to optimize the trade-off between treatment efficacy and drug induced toxicity during isoniazid chemotherapy. To this end, the developed PBPK/PD model was used to evaluate various isoniazid dose regimes in NAT2 specific immune competent and immune deficient virtual populations. Our simulation results suggest substantial dose adjustments for all acetylator types. We found bisecting the standard treatment for slow acetylators would substantially reduce the probability of adverse effects during isoniazid therapy, while maintaining high treatment efficacies. Especially, intermediate and fast acetylators would benefit from a combination of increased isoniazid doses and switching to a bi-daily administration regimen resulting in reduced treatment failures without serious gain of drug induced side effects. 103 Program / Abstracts PP 6-09 Validation of a pregnancy population physiologically-based pharmacokinetic model for renally cleared drugs Dallmann, A.1, Ince, I. 2, Meyer, M. 2, Eissing, T. 2, Hempel, G.1 University of Münster, Münster, Germany Bayer Technology Services, Systems Pharmacology CV, Leverkusen, Germany 1 2 Main objective: To validate a physiologically-based pharmacokinetic (PBPK) model for the prediction of pharmacokinetics (PK) of renally cleared drugs in populations of pregnant women at different stages of pregnancy. Methods: Based on a recent literature review on anatomical and physiological changes during pregnancy [1], a pregnancy population PBPK model has been developed using PK-Sim®/MoBi® [2]. In this model, the standard model structure of an adult woman was extended by 9 physiological compartments. These compartments are either specific to pregnancy (e.g. placenta and fetus) or become of specific relevance during pregnancy (uterus and breasts). The impact of the model structure extension was evaluated by comparing the simulated PK of virtual compounds using the standard model and the pregnancy model at pregnancy day zero. To ensure smooth transition from the non-pregnant to pregnant state, organs and blood flows at the onset of pregnancy were scaled to the non-pregnant levels. Populations of pregnant women were created using the organ scaling approach implemented in PK-Sim® [3]. The pregnancy population PBPK model was applied to predict the PK of predominantly renally cleared drugs at different stages of pregnancy. Prediction results were evaluated by comparison with experimentally observed literature data. Results: The extension of the model structure had only a negligible effect on the simulated PK of pregnant women at pregnancy day zero. The pregnancy population PBPK model successfully predicted the PK of all drugs at different stages of pregnancy. Compared to the non-pregnant state, maximum clearance changes of these drugs were observed in the early 2nd trimester with an increase of approximately 50%. The differences declined towards delivery, approximating values comparable to non-pregnant clearance levels. No changes in the activity of renal transporters involved in the clearance were necessary to correctly predict the experimentally observed PK. This indicates that the activity of these transporters remains essentially constant throughout pregnancy. Conclusion: We successfully developed and validated our pregnancy population PBPK model at different stages of gestation for predominantly renally cleared drugs. PK changes in pregnant women could be fully attributed to pregnancy-related changes in relevant parameters such as kidney volume and perfusion. Ultimately, this model can be applied to investigate in silico the PK of renally cleared drugs and help design clinical trials in this vulnerable special population. References: [1] Dallmann, A., 2015, PAGE 24, Abstr 3456. [2] Eissing, T. et al. (2011). A computational systems biology software platform for multiscale modeling and simulation: integrating whole-body physiology, disease biology, and molecular reaction networks. Frontiers in physiology 2: 1-10. [3] Willmann, S. et al. (2007). Development of a physiology-based whole-body population model for assessing the influence of individual variability on the pharmacokinetics of drugs. Journal of pharmacokinetics and pharmacodynamics 34(3): 401-431. PP 6-10 A hypothesis-free, systemic approach to identify pharmacological targets for anxiety disorders Nussbaumer, M.1, Maccarrone, G.1, Teplytska, L.1, Murphy, M. 2, Asara, J. 3, Landgraf, R.1, Turck, C.1, Filiou, M.1 Max Planck Institute of Psychiatry, Munich, Germany MRC Mitochondrial Biology Unit, Cambridge, Great Britain 3 Harvard Medical School, Boston, United States 1 2 Main Objectives: Anxiety disorders are the most prevalent psychiatric diseases worldwide. Since a third of the patients do not respond to existing anxiolytic treatments, hypothesis-free approaches to identify novel drug targets are necessary for improved response rates and personalized therapeutics. Materials and Methods: We developed a hypothesis-free, systems biology platform based on quantitative proteomics (15N metabolic labeling and mass spectrometry), metabolomics and bioinformatics to identify differences in mice selectively bred for high vs. low anxiety. We then pharmacologically manipulated the identified differences in high anxiety mice in vivo with appropriate compounds to assess whether selective pharmacological targeting exerts an alleviating effect in high anxiety. Results: Our hypothesis-free approach revealed altered brain mitochondrial pathways in high vs. low anxiety mice, including oxidative phosphorylation, oxidative stress and mitochondrial import/transport. These mitochondrial pathways were selectively manipulated in vivo by MitoQ, a mitochondria-targeted antioxidant. We observed a decreased anxietyrelated behavior in MitoQ-treated compared to untreated high anxiety mice. We then analyzed the molecular correlates of this anxiolytic effect with targeted metabolomics, immunoassays and biochemical assays. Conclusion: This is the first time that a hypothesis-free, mechanism- rather than symptom-based approach is used to 104 manipulate a behavioral phenotype. Our findings emphasize the potential of systems biology-driven approaches to discover new pharmacological treatments and reveal the therapeutic potential of mitochondrial targeting for brain disorders. PP 6-11 The NormSys registry for modeling standards in systems and synthetic biology Golebiewski, M.1, Nikolaew, A.1, Woetzel, N.1, Zander, J.1, Hollmann, S. 2, Mueller-Roeber, B. 2, Regierer, B. 3 HITS gGmbH, Heidelberg, Germany University of Potsdam, Focus Area of Plant Genomics and Systems Biology, Potsdam, Germany 3 LifeGlimmer GmbH, Berlin, Germany 1 2 The rapid development of modern life science technologies allows data generation with increasing speed and complexity. In systems biology these data have to be stored, shared, processed, integrated, analyzed and compared. Hence, standards for formatting and describing experimental data, applied workflows and resulting computer models have become a critical issue. Different stakeholders need to be engaged in the standardization process to incorporate their specific requirements: Researchers form academia and industries with their grass-roots standardization communities like the Computational Modeling in Biology Network (COMBINE: http://www.co.mbine.org), as well as representatives of standardization bodies (e.g. German DIN, European CEN/CENELEC or the International Organization for Standardization ISO), scientific journals and research funding agencies. The project NormSys, funded by the German Federal Ministry for Economic Affairs and Energy (BMWi) aims at enhancing and promoting the formal standardization of existing modeling community standards by building a bridge between stakeholder groups and developing the means for transferring information about community standards between them. To survey standard formats for computational modeling in systems biology and related fields we have developed the webbased NormSys registry for modeling standards (http://normsys.h-its.org). It provides a single access point for consistent information about model exchange formats such as the Systems Biology Markup Language (SBML), CellML, the Systems Biology Graphical Notation (SBGN), the Simulation Experiment Description Markup Language (SED-ML), the Synthetic Biology Open Language (SBOL), NeuroML for neuroscience models, the Pharmacometrics Markup Language (PharmML) and others. The publicly available platform not only lists the standards, but also compares their major features, their possible fields of biological application and potential use cases (including model examples), as well as their relationships, commonalities and differences. This NormSys registry provides a common entry point for anyone interested in modeling standards, especially for experimentalists, modelers and software developers who plan to apply the standard formats for their respective case of application, and serves them with detailed information, as well as with links to the webpages, specifications and web services of the formats. PP 6-12 Systems biology of MHC class I antigen presentation studied in human cancer cell lines Hahlbrock, J.1, Navarro, P.1, Wagner, M. 2, Boegel, S. 2, Schild, H.1, Sahin, U. 2, Tenzer, S.1 1 2 Institute for Immunology, University Medical Center Mainz, Mainz, Germany TRON gGmbH-Translational Oncology, University Medical Center Mainz, Mainz, Germany Question: Cancer is a primary cause of mortality in industrialized countries. Due to molecular heterogeneity in cancer, often less than 25% of treated individuals profit from the approved therapies. Individualized medicine is regarded as a potential solution to low efficacies and high costs for innovation in drug development and health systems, but requires comprehensive 105 Program / Abstracts data from individual samples and a well understanding of the underlying cancer biology. In the present study, we analyzed the proteome, transcriptome and ligandome of ten MHC class I matched human cancer cell lines derived from six different tissues with the aim to establish a prediction tool for MHC-presented peptide ligands, required for an antigen-specific immunotherapy. Methods: Tryptic peptides, generated from tumor cell lysates using a modified FASP protocol, were analysed by LC-MS using a nanoAcquity UPLC system coupled to a Waters Synapt G2-S HDMS instrument. Rawdata analysis was performed in PLGS3.02, searching UniProtKB/Swissprot (human reference proteome). Final identification and clustering steps were performed with ISOQuant. RNA analyses were performed as a paired-end sequencing of cell line cDNA (2x50 nt) with an Illumina HiSeq 2500 system. Reads were mapped against the human genome (hg19) using the STAR mapping Algorithm. MHC class I molecule associated peptides (ligandome) were isolated using a modified immunoaffinity purification protocol binding the molecules to W6/32 coupled BrCN-Sepharose beads. Results: Using a novel data-independent acquisition (DIA) label-free LC-MS approach for proteome analysis, we established a reference proteome data set for those cancer cell lines covering in total over 6.700 protein groups and their relative abundances. On mRNA level we identified between 9.791 and 12.467 expressed (>= 1 RPKM) coding genes. Analyzing correlations between proteomics and transcriptomics data, we found surprisingly high correlation values, ranging from r= 0.5 to 0.58 (spearman correlation coefficient). Gene Ontology analysis (protein folding, gene expression, RNA binding, etc.), revealed correlation values up to r= 0.89. For ligandome analyses we identified between 762 and 3.338 ligands, whereby the number of identified ligands correlated with the respective MHC expression level. Conclusion: We generated comprehensive proteomic, transcriptomic and ligandomic data sets, which will be integrated into a model to quantitatively predict the MHC class I ligandome and to establish novel strategies for antigen-specific immunotherapy. PP 6-13 NormSys & CHARME Two initiatives that aim at Harmonizing the Standardization Processes for Data Exchange in Systems Biology Hollmann, S.1, Golebiewski, M. 2, Regierer, B. 3 University of Potsdam, Plantgenomics and Systems Biology, Potsdam, Germany HITS gGmbH, , Heidelberg, Germany 3 LifeGlimmer GmbH, Berlin, Germany 1 2 An essential prerequisite of modern life-science R&D is a high quality of the research data. By enabling the reuse of research assets, research becomes considerably more efficient and economical. In this context a common understanding of formats and standards for data and computer models particularly in interdisciplinary research fields is clearly evident: this encompasses detection and description of the data, their efficient and secure exchange between research labs, institutes and industry, their integration into workflows, as well as their processing and analysis. Thus, standards represent important drivers in the life-sciences and technology transfer because they guarantee that data become accessible, shareable and comparable along the value chain. Several initiatives launched the development and implementation of standards. Unfortunately these efforts remain fragmented and largely disconnected. The two initiatives NORMSYS & CHARME will merge the different approaches in the field with a particular reference to systems biology, and thus avoid too many different solutions being generated in parallel universes that “in the worst case” are neither compatible nor suitable for large-scale approaches. NormSys is a research project funded by the German Ministry of Economic Affairs and Energy (BMWi) aiming at enhancing and promoting the formal standardization of existing community standards in systems biology on a national level. It develops a concept for the transformation of these standards into officially certified specifications. NormSys establishes a platform for relevant stakeholders to participate in this standardization process for incorporation of their specific requirements: researchers from academia and industries with their grass-roots standardization communities like the Computational Modelling in Biology Network (COMBINE), as well as representatives of official standardization bodies (eg DIN, CEN-CENELEC or ISO), scientific journals and funding agencies. CHARME is a international project (COST Action) starting in Mai 2016 which aims to increase the awareness for the need of standards, enabling the reuse of research data and its interoperability within the scientific community. CHARME provide a common ground for researchers from academia, research institutes, SMEs and multinational organizations on an international level. Both of the above projects complement each other in an optimal way: they will 106 • firstly combine and review existing community standards and standardization options including the development of a common understanding/definition of the needs, • secondly push the implementation of minimal standards in biotechnology especially in systems biology, and • thirdly create a joint platform for stakeholders for sustainable and efficient exchange and cooperation PP 6-14 A dynamic model of bile acid transport in the HepaRG cell-line Kaschek, D.1, Sharanek, A. 2, Guillouzo, A. 2, Timmer, J.1,3,4, Weaver, R. J. 5 University of Freiburg, Institute of Physics, Freiburg, Germany Université de Rennes 1, Faculté des Sciences Pharmaceutiques et Biologiques, Rennes, France 3 University of Freiburg, BIOSS Centre for Biological Signalling Studies, Freiburg, Germany 4 University of Freiburg, Freiburg Center for Systems Biology (ZBSA), Freiburg, Germany 5 Institut de Recherches Internationales Servier (I.R.I.S.), Paris, France 1 2 Hepatobiliary transporters might play a major role in the development of drug-induced liver injury (DILI). In the past, various in vivo and in vitro studies have been carried out analyzing uptake and efflux of bile acids and the impact of cholestatic versus non-cholestatic drugs. Although human hepatocytes are considered the most realistic in vitro cell model, the huge variability observed in this system limits its suitability for the development of a quantitative mathematical model of bile acid transport. Here we present a dynamic model based on ordinary differential equations that describes uptake, basolateral export and canalicular export of taurocholic acid, a bile acid, in the HepaRG cell line. The highly reproducible experimental data obtained in HepaRG cells allows to estimate transport rates associated to different transporters like NTCP, MRP3 and BSEP. Experiments with the cholestatic drug cyclosporin identify the dose-dependent inhibition of single transporters. Finally, we compare our results in HepaRG to efflux-dynamics in human hepatocytes and determine the pivotal differences between both cell types, thus, allowing to predict bile transport in human hepatocytes based on a calibrated model in HepaRG. Funding: This work was supported by funding from the MIP-DILI project, a European Community grant under the Innovative Medicines Initiative (IMI) Programme (Grant Agreement number 115336). PP 6-15 Modeling Individual Time Courses of Thrombopoiesis During Multi-Cyclic Chemotherapy Kheifetz, Y.1, Scholz, M.1, Löffler, M.1 1 IMISE (Institut Für Medizinische Informatik, Statistik und Epidemiologie), Leipzig, Germany Question: Decreased platelet counts, called thrombocytopenia, is a major dose-limiting side effect of dose-intense cancer chemotherapies. However, standard coursers of many chemotherapies result in considerable variability in drug induced platelets dynamics. A major challenge of individualized medicine is to take all relevant factors into account for optimal risk management using individualized modeling. Actually there is a gap between simple phenomenological predictive population models and complex biomathematical modeling of averaged data. In order to tackle this problem we revised a biomathematical model of average human thrombopoiesis under chemotherapy (Scholz et al. 2010) towards modelling individual time courses. Methods: We fitted model parameters hierarchically. More than 20 parameters were estimated using dense time series of three patients treated with BEACOPP chemotherapy. We used these estimators as 3 alternative population models in order to fit individually 6 parameters for sparser data of selected 135 patients from the German non-Hodgkin‘s lymphoma trial group applying CHOP-like chemotherapies. The individual parameter estimations simultaneously used information from other studies published in literature by incorporating it into the respective weighted likelihood functions, assuming a virtual participation of the patients in the corresponding experiments. These additional information included average dynamics of TPO, platelets and megakaryocytes of recombinant-TPO treated healthy patients as well as a platelets dynamics after labeled platelets transfusions to patients with different degrees of thrombocytopenia. Result: Several new biological insights were discovered and modeled. We hypothesize a bi-phasic TPO-stimulation of thrombopoiesis. The slowly long-term decrease in average platelets level during multi-cyclic chemotherapy was attributed to interactions between quiescent and active stem cells compartments. We have found that multi-cyclic chemotherapy significantly reduces transit times for megakaryocytes and platelets. These long-term systems changes during multi-cyclic polychemotherapies are responsible for strong intra-individual variability of platelets’ nadirs and consequently of chemotoxicity through treatment cycles. We have found an optimal tradeoff between goodness of fit and overfitting for most of the patients. Conclusions: We established a model of individual thrombopoiesis response to chemotherapy. Heterogeneity between patients can be traced back to heterogeneity of a few model parameters. We will exploit the predictive potential of the model in the near future. 107 Program / Abstracts PP 6-16 DESIGNING AND SIMULATION STUDIES OF Mycobacterium Tuberculosis DNA GYRASE. Niranjan, V.1, Chaudhuri, T. 2, Prasanna, A.1, N. K. N2, Deshpande, S. 1 1 2 RVCE, BIOTECHNOLOGY, Bengaluru, India GenEclat Technologies, Bioinformatics, Bengaluru, India Tuberculosis is one of the airborne and contiguous diseases which ranks second leading causes of death that accounts for the increase in the mortality rate after HIV. The Mycobacterium tuberculosis is a causative organism for tuberculosis. Unlike any other bacteria, the M. tuberculosis genome analysis has identified a gyrB-gyrA contig in which gyrA and gyrB encode the A and B subunits of DNA Gyrase, respectively. In our present study, we have focused on different methods involved in designing the DNA Gyrase inhibitor for M. tuberculosis using the techniques like Molecular Docking and Molecular Dynamics using SCHRODINGER SUITE. MATERIALS and METHODS: PDB ID for target DNA Gyrase (Subunit A and B) was identified through http://www.rcsb. org/ and screening of PDB were done based on X-Ray resolution and the Scaffold structure of Ligand. The screened PDB’s was docked. The Post-Reproducibility was carried out by using MAESTRO. Homology Analysis was carried out by using PRIME. The protein sequence was retrieved from the UNIPROT database. For Docking study Chembl 2.0 ligand database was selected for studies based on the Number of molecules and the total number of experiments carried on Homo sapiens. It employs combinatorial technology for lead identification and optimization. Chemoinformatics analysis was carried out by using CANVAS 2.1. To perform molecular docking, the software GLIDE 6.0 was used. The High throughput Virtual Screening method was applied first to screen the ligand database consisting of 5,17,261 molecules. Hierarchical clustering was carried out for generation of Dendrogram. In silico screening of drug-likeness was done by Qikprop. Molecular Dynamics was carried out using DESMOND. The system was relaxed either by minimization or by selecting the panel option to relax the model system before simulation. The simulation parameters were set in one of the general Desmond panels. RESULT and DISCUSSION: In our present study, Homology Analysis was carried using PRIME SCHRODINGER SUITE, where two targets receptor i.e. 3ZM7.pdb (DNA Gyrase A) and 3IFZ.pdb (DNA Gyrase B) showed a better result. These two structures showed a high resolution(R value of 3.30 A° and 2.70 A° respectively) when downloaded from PDB. Hence, they were subjected to docking study against Chembl 2.0, a tool used that generates a combinatorial library. The Cluster 2 which had three poses had a best Average score of -8.667. Protein-ligand interactions were explored through the ‹Simulation Interactions Diagram› panel. CONCLUSION: The M. tuberculosis genome analysis has identified a gyrB-gyrA contig in which gyrA and gyrB encode the A and B subunits. Newer fluoroquinolones, including Moxifloxacin and Gatifloxacin, exhibit potent activity against M. tuberculosis and show potential to shorten the duration of TB treatment. M. tuberculosis DNA Gyrase is thus a validated target for anti-tubercular drug discovery. A novel inhibitor of M. tuberculosis DNA Gyrase would be effective against multidrug resistant (MDR)-TB, and it could also be effective against Fluoroquinolones-resistant M. tuberculosis. PP 6-17 Early biomarkers and magnetic resonance imaging for diagnosing Bronchopulmonary Dysplasia Sass, S.1, Förster, K. 2,3, Theis, F.4,1, Hilgendorff, A. 2,3 Helmholtz Zentrum München, Institute of Computational Biology, Neuherberg, Germany Dr. von Hauner Children’s Hospital, Dept. of Neonatology, Munich, Germany 3 Helmholtz Zentrum München, Comprehensive Pneumology Center, Munich, Germany 4 Technical University of Munich, Department of Mathematics, Grosshadern, Germany 1 2 Rationale: Neonatal chronic lung disease, i.e. BPD, as a consequence of pre- and postnatal injury to an immature lung determines long-term pulmonary and neurologic development. As to date, the diagnosis is solely stated with respect to its clinical course referring to oxygen dependency on day 28 post partum or at 36 weeks postmenstrual age (PMA). Early markers are urgently needed for timely diagnosis and personalized treatment. Objectives: In a unique approach this prospective study determined structural and functional changes in the preterm lung at the time of diagnosis and identified early disease markers by proteome screening in plasma in the first week of life. Method: Forty preterm infants (27.7±2.09wks, 984±332g) were included for advanced MR imaging (3-Tesla) and complemented by Infant lung function testing (ILFT) in spontaneously breathing infants. Plasma samples were processed for proteomic screening by SOMAscan™. Key findings were confirmed in an independent study cohort (n=21 infants). Statistical analysis used penalized and Poisson regression analysis; for protein analysis confounder effects were subtracted by lasso regression. Results: Statistical analysis confirmed a high correlation of MRI and lung function variables and identified a pattern characterizing changes in the lungs of preterm infants by T2- and T1-weighed image analysis and lung volume measurements as well as ILFT. Statistical modelling using the outcome variables ‘days of oxygen’ or ‘days of MV’ instead of BPD diagnosis 108 confirmed the analysis indicated above. Functional enrichment analysis showed overrepresentation of the GO categories ‘immune function’, ‘extracellular matrix’, ‘cellular proliferation/migration’, ‘organ development’ and ‘angiogenesis’ in infants with BPD. Second, 12 proteins were significantly regulated when comparing the group of infants later developing BPD with infants without the disease, reflecting the categories identified by functional enrichment analysis. One protein was identified as a potential biomarker. Conclusions: We identified a structural pattern characterizing BPD by advanced MRI confirmed by ILFT. The findings reflect the characteristic picture outlined by previous studies of human tissue samples, i.e. the presence of interstitial and emphysematous changes. The identified protein indicated BPD development in the first week of life enabling personalized treatment strategies. PP 6-18 Using physiologically-based pharmacokinetic modelling to analyze the effect of hepatic impairment on drug detoxification capacity at the whole-body level Schenk, A.1,2, Ghallab, A. 3,4, Hassan, R. 3,4, Hofmann, U. 5,6, Schuppert, A.1,7,2, Teutonico, D.7, Hengstler, J. 3, Kuepfer, L.7 RWTH Aachen, Joint Research Center for Computational Biomedicine, Aachen, Germany RWTH Aachen, Aachen Institute for Advanced Study in Computational Engineering Science (AICES), Aachen, Germany 3 Leibniz Research Centre for Working Environment and Human Factors at the Technical University Dortmund, Dortmund, Germany 4 South Valley University, Department of Forensic Medicine and Toxicology, Faculty of Veterinary Medicine, Qena, Germany 5 Dr. Margarete Fischer-Bosch Institute for Clinical Pharmacology, Stuttgart, Germany 6 University of Tuebingen, Tuebingen, Germany 7 Bayer Technology Services GmbH, Computational Systems Biology, Leverkusen, Germany 1 2 Main objective: To quantify and analyze the effect of hepatic impairment on pharmacokinetics of a drug cocktail at the whole-body level by means of physiologically-based pharmacokinetic (PBPK) modelling. Methods: For healthy mice, we measured the plasma concentration profiles of a drug cocktail of six parent drugs (caffeine, codeine, midazolam, pravastatin, talinolol, torsemide) including some of the corresponding metabolites. Notably, we considered the plasma concentration profiles at three sampling sites within the body, i.e. the heart, the portal vein and the hepatic vein. Additionally, we quantified the area of expression within the liver of selected cytochrome p450 enzymes by immunostaining. We used this data set to firstly establish PBPK models for the six compounds for healthy mice including in particular liver zonation with a periportal and a pericentral region based on the results of the stainings. Notably, the mechanistic nature of the PBPK models allowed us to consider all three sampling sites at the same time. Next we quantified the impact of CCl4 -induced liver damage on the hepatic metabolic capacity. To this end, the same plasma concentration profiles data was also gathered for mice after CCl4 -administration. Results: The six PBPK models for the healthy mice were in excellent agreement with the experimental concentration profiles at all three sampling sites as such allowing a quantification of hepatic metabolic capacity at the whole-body level. Notably, the PBPK models include experimental information from different scales of biological organization ranging from enzyme expression in the hepatic lobule to drug concentration profiles in the blood plasma. Furthermore, it was possible to describe the plasma concentration profiles of mice after CCl4 -induced liver damage. Conclusion: We successfully established PBPK models for a cocktail of six drugs for healthy mice. These models feature liver zonation and concentration profiles from three different sampling sites within the body. The approach allows the quantification and analysis of hepatic impairment at the whole-body level. PP 6-19 Dynamical modelling of the murine immune response to pneumococcal lung infection with and without antibiotic treatment Kuepfer, S.1, Ahnert, P.1, Loeffler, M.1, Wienhold, S. 2, Nouailles-Kursar, G. 2, Witzenrath, M. 2, Scholz, M.1 University of Leipzig, IMISE, Leipzig, Germany Charite Universitaetsmedizin, Berlin, Germany 1 2 Pneumonia is considered to be one of the leading causes of death worldwide. The outcome depends on both, proper antibiotic treatment and the effectivity of the immune response of the host. However, due to the complexity of the immunologic cascade initiated during infection, the latter cannot be predicted easily. We construct a biomathematical model of the murine immune response during infection with pneumococcus aiming at predicting the outcome of antibiotic treatment. The model consists on a number of non-linear ordinary differential equations describing dynamics of pneumococcal population, the inflammatory cytokine IL-6, neutrophils and macrophages fighting the infection and destruction of alveolar tissue due to pneumococcus. Equations were derived by translating known biological mechanisms and assuming certain response kinetics. Antibiotic therapy is modelled by a transient depletion of bacteria. Unknown model parameters were determined by fitting the predictions of the model to data sets derived from mice 109 Program / Abstracts experiments of pneumococcal lung infection with and without antibiotic treatment. Time series of pneumococcal population, debris, neutrophiles, activated epithelial cells, macrophages, monocytes and IL-6 serum concentrations were available for this purpose. The antibiotics Ampicillin and Moxifloxacin were considered. Parameter fittings resulted in a good agreement of model and data for all experimental scenarios. Sensitivities of parameter estimates could be estimated. The model can be used to predict the performance of alternative schedules of antibiotic treatment. We conclude that we established a biomathematical model of pneumococcal lung infection in mice allowing predictions regarding the outcome of different schedules of antibiotic treatment. We aim at translating the model to the human situation in the near future. PP 6-20 Modelling immuno-chemotherapy of lymphomas Scholz, M.1, Roesch, K.1, Hasenclever, D.1 1 IMISE / University of Leipzig, Leipzig, Germany Background: Moderate intensifications improved the outcome of lymphoma chemotherapy, but highly intense therapies are inferior. We hypothesise that the immune system has a key role in controlling residual tumour cells after treatment. More intense therapies result in a stronger depletion of immune cells allowing an early re-growth of the tumour. Methods: To understand this process in more detail, we propose a differential equations based model of the dynamics and interactions of tumour and immune cells under chemotherapy. Major model features are an exponential tumour growth, a modulation of the production rate of effector cells by the presence of the tumour (immunogenicity) and mutual destruction of tumour and immune cells. Chemotherapy causes damage to both, immune and tumour cells. Immunotherapy by the monoclonal antibody Rituximab is modelled by a direct cell kill and an intensified immune response. Growth rate, chemosensitivity, immunogenicity and initial size of the tumour are assumed to be patient-specific, resulting in heterogeneity regarding therapy outcome. Maximum-entropy distributions of these parameters were estimated on the basis of clinical survival data. Results: The resulting model can explain the outcome of eight different chemotherapeutic regimens with and without Rituximab and corresponding hazard-ratios. Estimated parameters are biologically plausible. We demonstrate how the model can be used to make predictions regarding yet untested therapy options. Conclusions: We conclude that our model explains observed paradox effects in lymphoma therapy by the simple assumption of a relevant anti-tumour effect of the immune system. Heterogeneity of therapy outcomes can be traced back to heterogeneity of a few model parameters whose distribution can be estimated on the basis of clinical survival data. The model can be used to predict the performance of new therapy options. PP 6-21 Nuclear-encoded mitochondrial genes in hypercholesterolema and atherosclerosis progression Vilne, B.1, Björkegren, J. L. 2, Skogsberg, J. 3, Foroughi, Asl H. 3, Kessler, T.1, Schunkert, H.1 Deutsches Herzzentrum München, Klinik für Herz- und Kreislauferkrankungen, München, Germany Icahn School of Medicine at Mount Sinai, Department of Genetics and Genomic Sciences Icahn Institute for Genomics and Multiscale Biology, New York, United States 3 Karolinska Institutet, Cardiovascular Genomics Group, Division of Vascular Biology, Department of Medical Biochemistry and Biophysics, Stockholm, Sweden 1 2 Background and Aims: Dysregulation of nuclear-encoded mitochondrial gene (NEMG) expression might constitute an intermediate step by which hypercholesterolemia promotes atherosclerotic lesion development. Methods and Results: The transcriptome of the aorta was examined in Ldlr−/−Apob100/100 Mttpflox/flox Mx1-Cre mice at 10-week intervals. During the phase of rapid lesion expansion and foam cell formation, i.e. between 30 and 40 weeks of age, NEMGs were 4-fold down-regulated compared to non-NEMGs (P = 6.11e-72) and involved >28% of total NEMGs (n=1,240). Down-regulation included PGC1-α and -β, i.e. two members of the peroxisome proliferator-activated receptor γ co-activator 1 family of transcription regulators. In addition, we identified two mitochondria-related gene co-expression modules as being the most negatively associated with lesion development (r= -0.42 and -0.43; P=0.04). Within these modules, we observed significant over-representation of estrogen related receptor ERR-α binding sites, reactive oxygen species (ROS) target genes and several key mitochondrial pathways. In fact, the expression of ERR-α, an effector of PGC1 in the control of mitochondria biogenesis, was found tightly correlated with PGC1, mitofusin 2 (MFN2), uncoupling protein 3 (UCP3), superoxide dismutase 2 (SOD2), lipin 1 (LPIN1) and citrate synthase (CS) expressions. Plasma cholesterol lowering (PCL) at 30 weeks, previously shown to prevent advanced plaque formation, resulted in 4-fold up-regulation of NEMGs (P = 3.89e-16). 110 Conclusion: Our results suggest that arterial wall responds to prolonged hypercholesterolemia by a co-ordinated downregulation of NEMGs, including mitochondrial biogenesis and antioxidant genes, possibly regulated by ERR-α/PGC1. Downregulation demonstrates an inverse relationship to lesion expansion and foam cell accumulation, and is partly reversed by PCL. PP 7: Systems Medicine & Genetic/ Epigenetic Mechanisms PP 7-01 Acute and Chronic inflammation and mast cell activation, IN Silico Abdukerim, A.1, Barberis, M.1, Sahin, N. 2, van Ham, M.1,3, Westerhoff, H. V.1,2,4 University of Amsterdam, Synthetic Systems Biology , Amsterdam, Great Britain VU University Amsterdam, Department of Molecular Cell Physiology, Amsterdam, Great Britain 3 Sanquin Blood Supply, Department of Immunopathology, Amsterdam, Netherlands 4 Manchester Interdisciplinary Biocentre, Manchester Centre for Integrative Systems Biology, Manchester, Great Britain 1 2 Main objectives of the study: Systems Immunology addresses how regulators of the immune system that integrates at various system levels, modulate the response to inflammation. We aimed to design a computer model able to provide a rational explanation of the network’s response to antigen in terms of acute or chronic inflammation. The model should calculate in silico the level of tumor necrosis factor alpha (TNF-α). It should likewise predict the effect of an in silico drug targeting the immunoglobulin free light chain (FLC) that activates mast cells, on the outcome of an inflammation. Materials and methods: A network of the native immune system was decomposed into various dynamic processes, the activation of which depending on the concentration of their components in time. For each species in the network, a balance equation was then formulated, specifying its time dependence as the difference between the rates of the reactions synthesizing the commodity and the rates of the processes degrading it. The corresponding rate equations were integrated in time or for steady state by using the COPASI simulation software. Immunostaining on cancer biopsies, protein detection, and in vitro and in vivo studies were used for validation. Results: The model enabled us to 1) examine scenarios such as the capability of the system to kill invading microorganisms, 2) explain the dichotomy between acute and chronic inflammation, and 3) test the effect of a in silico peptidic drug that inhibits FLC. In our experimental studies FLCs secreted from B-cells bound to their specific receptor on mast cells.FLC and mast cells were co-localized in various cancer biopsies such as those of lung cancer tissues. Therapeutically, administration of an anti-FLC peptide had a beneficial effect on tumorigenesis, suggesting that FLC may play a role by targeting inflammation that promotes cancer. The model reproduced these findings and was subjected to tests for dynamic stability and to sensitivity analysis. Conclusions: An iterative process of model building is able to provide new understanding about the mechanisms of inflammation. A systems mechanism for the effects of therapeutic peptides against inflammation and tumorigenesis was identified and made computable in terms of personalized molecular properties. This should facilitate further testing as well as provide a platform for personalized systems immunology. PP 7-02 Prioritization of candidate causal genes in rare genetic diseases through integration of exome sequencing data and biological databases Avsec, Z.1, Mertes, C.1, Prokisch, H. 2, Gagneur, J.1 TU Munich, Bioinformatics, Garching bei München, Germany Helmholtz Zentrum München, Institut für Humangenetik, Neuherberg, Germany 1 2 Identification of novel causal genes in rare Mendelian diseases helps to understand the disease’s pathomechanism, solve more cases and reveal novel biological pathways. For proving the causality of a gene, experimental follow-up studies are needed. These are usually expensive and time-consuming, hence the genes sent for validation should be chosen carefully. Here, we propose different statistical models for prioritizing candidate genes that integrate exome sequencing data, known causal genes, molecular pathways and protein-protein interaction networks. The models were applied to a whole-exome sequencing dataset of about 550 mostly unrelated cases diagnosed with a mitochondrial disease and 300 controls. Sensitivity for recovering 75 known causal genes in a cross-validated fashion was used to benchmark the models. Bayesian approaches showed increased sensitivity over classical statistical testing. Highest sensitivity -40% in the 100 most highly ranked predictions- was reached using the machine learning approach. Our method is general and could be applied to other rare diseases with a few known causal genes. 111 Program / Abstracts PP 7-03 A glance at smoking effect on preeclamptic placenta with transcriptional regulation approach Aydin, B.1, Arga, K.1 1 Marmara University, Bioengineering, Istanbul, Turkey Preeclampsia (PE) is a pregnancy specific disease that indicated by glitches of the placenta like abruptions or infarctions. PE can cause adverse pregnancy outcomes like prematurity and growth restriction of baby meanwhile hypertension and proteinuria for mother. Smoking affects baby by plecanta previa, low birth weight but paradoxically risk of PE is lower in smoker mothers than non-smoker ones. Exact mechanism and relation between cigarette smoking and preeclampsia has not been enlightened yet. Objective: To search whether a preventive side of smoking on preeclampsia. If so, to investigate a relation between cigarette smoking and preeclampsia by using transcription factors (TFs). Methods: To elucidate gene expression differences, three high throughput screening datasets downloaded (GSE44711preeclamptic plancenta, GSE7434- cigarette smoking placentas, GSE48424 smoking-preeclamptic placenta). Differentally expressed genes (DEGs) are sorted out and transcription factors that regulates these DEGs are determined. A program TRANSREGNET coded which includes 57157 genes and 348 TFs that allows the user finding of Tfs which is responsible for regulation of query genes. By using this program Tfs were found for DEGs from each three cases. Then these TFs were classified by their job. Results: Preeclamptic placenta, cigarette smoking placenta and smoking-preeclamptic placenta samples share common TFs of AR, E2F4, ESR1, FOXA1, FOXP3, GATA1, GATA2, GATA3, MYC and YBX1. Among these TFs ESR1 and GATA3 regulates uterus development, MYC is responsible from Wnt signaling which is essential for female reproduction, uterine function and organ development. ESR1 and E2F4 have roles in epithelial cell development which is important for PE. Because epithelial dysfunction plays an essential role in pathogenesis of PE, also. Conclusion: The system behind cigarette smoking in pregnancy and reduced PE risk is not fully clear in literature yet. While the hierarchy between genes and cross-talk of TFs in our model, there could be a logical relation between smoking during pregnancy, lowers PE risk by regulation of oxidative stress and prevention of vascularization. PP 7-04 Post-transcriptional gene regulation in mitochondrial disorder patients Bader, D. M.1, Kremer, L. S. 2, Pichler, G. 3, Schwarzmayr, T. 2, Holzerova, E. 2, Kopajtich, R. 2, Wieland, T. 2, Strom, T. M. 2,4, Mann, M. 3, Prokisch, H. 2,4, Gagneur, J.1 Technische Universitaet Muenchen, I12 Computational Genomics, Munich, France Helmholtz Zentrum, Institute for Human Genetics, Neuherberg, Germany 3 Max-Planck Institute of Biochemistry, Department of Proteomics and Signal Transduction, Martinsried, Germany 4 Technische Universitaet Muenchen, Institute of Human Genetics, Munich, Germany 1 2 Joint analysis of transcriptome and proteome allows a more detailed explanation of the observed phenotype and the regulatory steps that lead to its intermediates, than separate studies of RNA and protein levels. Recently, researchers started integrative genome-wide studies on RNA and protein levels. Yet, most effort was spent on model organisms leaving the complex regulation of protein per RNA (PPR) in humans largely unexplored. Here, we investigate PPR regulation in 35 human fibroblast cell lines, originating from mitochondrial disease patients. Our PPR analysis is based on the comparison of more than 4,500 gene products measured with RNAseq and mass-spectrometry. We demonstrate that controlling for hidden technical biases in the data rescues the similarity of biological replicates. Interestingly, pathways showed various levels of RNA protein correlation with the ribosome complex showing significant low correlations and sarcolemma high correlations. This indicates that pathways adopt different regulatory strategies, which are rather transcriptional or post-transcriptional. Additionally usage of frequent codons correlates with the PPR expression ratio, consistent with the positive contribution of frequent codons to translation rates. The joint analysis of RNA and protein in human cell lines widens our view on cellular architecture and regulation. 112 PP 7-05 Quantitative analysis of gene expression based on large distance histone acetylation modifications as enhancers Haghighi, E.1, Heinig, M.1 1 Helmholtz Zentrum Muenchen, Institute of Computational Biology, Neuherberg, Germany Objectives: Enhancers are highly tissue specific and important for the regulation of transcription of their target genes. Previously it has been shown that levels of histone modifications in transcription start sites are highly predictive of gene expression levels. Here we aim to study the role of long distance enhancers for tissue specific gene expression using quantitative analysis of gene expression based on the levels of the histone modification H3K27ac. Methods and materials: We used gene expression and histone modification data for 19709 genes and 48 tissues from the epigenomics roadmap consortium. In order to predict the gene expression level, a linear model which is penalized using Lasso is fitted on the different sites of H3K27ac, for which the number of reads per kilo base per million (RPKM) is computed. The evaluation process is accomplished using one-leave-out method. For each gene the following steps are applied: once a tissue is left out, the K nearest tissues are considered to learn a linear model to estimate their gene expression with respect to the relevant histone acetylation sites. These relevant sites are selected based on the most significant correlations between the measured RPKM for gene expression and H3K27ac sites. Using the estimated model, the gene expression can be predicted for the excluded tissue. It is also worth reminding that the nearest tissues are determined once using the distance between gene expression vectors for each pair of tissues over the whole genome as a preprocessing procedure. Results: The proposed method is evaluated for each gene based on the correlation between the actual level of the gene expression and the estimated one. The proper value of K is a trade off point between selecting more similar tissues and most relevant acetylation sites (lower values of K) and being sure to have enough information for estimation (higher values of K). With respect to our experiments, the best value to set K is 8, for which the correlation of the results for about half of the genes is more than 0.5. The proposed method is more effective for the genes, which are active in a larger number of tissues. For the genes with more than 20 active tissues, the mean of the correlation is 0.49, whilst for the overall genes is 0.44. Conclusion: In this abstract, a linear model to predict gene expression based on long distance enhancers is presented. Applying a proper value for the number of K nearest tissues, is important to use the relevant information to estimate the model. For about half of the genes, the estimated models correlates reasonably with the actual values of gene expression. The ability of predicting gene expression level increases for the genes, which are active in most of the tissues. Our results also allow to associate long distance enhancers to their target genes and can be used for the interpretation of the genetic variations related to cancer or common diseases. PP 7-06 Time series analysis of macrophage activation and signalling from RNAseq data Bergey, F.1, England, H. 2, Papoutsopoulou, S. 2, Werner, M. 2, Paszek, P. 2, Kittner, M.1 1 2 LifeGlimmer GmbH, Berlin, Germany University of Manchester, Faculty of Life Science, Manchester, Great Britain Time series high-throughput data require advanced procedures to compare between different parameters. Here we developed new statistical methods based on linear regression models to determine temporally regulated gene clusters in macrophage cells undergoing differential immune activation. The RNA-seq data used in this work have been generated within the SysmedIBD consortium (FP7, grant agreement n° 305564). Macrophages are our first line of defence against foreign threats, but their activation must be coordinated to avoid harmful effects. Here we investigated their temporal response of mouse bone marrow derived macrophages to the pro-inflammatory cytokine tumour necrosis factor alpha (TNF-α), and lipid A, agonist of toll-like receptor signalling, a major pathogen sensing system. Samples were collected for the untreated condition and 1 hour, 3 hours and 6 hours after treatment. A linear model is estimated based on the gene expression values using time, treatment and the interaction between time and treatment as explanatory variables. Statistical tests were performed to estimate the role of these variables between treatments. Based on this analysis, genes were classified depending on their response pattern after treatment. We focused on some particular, temporarily regulated gene clusters such as genes activated early after stimulation and then switched off, or genes activated early or late with prolonged expression. In these specific clusters, we analysed whether the treatments induce different response patterns or different expression levels, and established treatment specific effects. For instance, A20 and IκBα genes, negative regulators of the overall response, belong to the first cluster and show similar activation patterns for both treatments. However, key signalling molecules such as interleukin 1α and β, show differential activation: TNF-α induces early transient activation whereas lipidA causes a sustained response. Genes within clusters were then further analysed for molecular functions and interactions on the protein level. The linear model designed in this study enables grouping of genes according to their significant response patterns and easily determines genes that respond differently to one of the treatments. Such an approach can be applied similarly to determine differences in expression when comparing different cell types or between species. 113 Program / Abstracts PP 7-07 In situ transcriptomics reveals metabolic and immunological zonation in human liver Brosch, M.1, Muders, M. 2, Matz-Soja, M. 3, Schafmayer, C.4, von Schönfels, W.4, Zeissig, S.1, Dahl, A. 5, Segiova, F.6, Zerial, M.6, Küpfer, L.7, Brusch, L. 8, Kaderali, L. 9, Gebhardt, R. 3, Becker, T.4, Baretton, G. B. 2, Ehninger, G.1, Hampe, J.1 University Hospital Carl Gustav Carus, Department of Internal Medicine I, Dresden, Germany University Hospital Carl Gustav Carus, Institute of Pathology, Dresden, Germany 3 University of Leipzig, Institute of Biochemistry, Leipzig, Germany 4 University Hospital Schleswig-Holstein, Department of General and Thoracic Surgery, Kiel, Germany 5 Technische Universität Dresden, BIOTEChnologisches Zentrum der TU Dresden, Dresden, Germany 6 Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany 7 Bayer Technology Services GmbH, Computational Systems Biology, Leverkusen, Germany 8 Technische Universität Dresden, Center for Information Services and High Performance Computing, and Center for Advancing Electronics Dresden, Dresden, Germany 9 University Medicine Greifswald , Institute for Bioinformatics,, Greifswald, Germany 1 2 Liver parenchyma shows a considerable and functional plasticity, known as metabolic zonation. Thus, different metabolic pathways are carried out in different, the periportal and pericentral, zones of the liver lobules. Along the axis of the lobule all liver functions need to be performed resulting in a heterogeneous distribution of enzymes, transporters and other functional components. At least for the mouse a pattern of gene expression could be drawn on the basis of microarray gene expression analysis on preparations of isolated periportal and pericentral hepatocytes. To investigate the global gene expression pattern in human liver zonation we used RNA-Seq on laser captured microdissected periportal and pericentral human hepatocytes. In total 261 genes were identified, that showed significant changes (p<0.01) between the hepatocytes of the two different zones of the liver. The results are in good concordance with the described gene expression in mice with a differential expression of genes encoding enzymes of the intermediary metabolism, ammonia utilization, amino acid metabolism and xenobiotic metabolism. In addition genes of proteins of the zonal driver pathways Wnt/b-Catenin and Hedgehog were complementary expressed in periportal and pericentral hepatocytes. Furthermore an innate immunity zonation was observed with a high expression of genes encoding for antimicrobial proteins in hepatocytes surrounding the periportal field. Immunohistochemical staining validated the spatial distribution of c-reactive protein. In conclusion, the human data confirms the metabolic zonation as it is observed in mouse expression analysis, indicating that whole tissue approaches have to be assessed carefully under consideration of the heterogeneous and functional plasticity of hepatocytes along the liver lobule axis. Furthermore the periportal expression of genes encoding for proteins of the innate immune system indicates that the hepatocytes surrounding the periportal field might function as a barrier for gut derived pathogens. PP 7-08 Deep learning of regulatory factors of the transcriptional landscape Eraslan, G.1, Preusse, M.1, Theis, F. J.1,2, Mueller, N. S.1 1 2 Helmholtz Zentrum München, Institute of Computational Biology, Neuherberg, Germany Technische Universität München, Department of Mathematics, Garching, Germany Myriad of RNA-binding proteins (RBP) and ribonucleoproteins (RNP) are involved in many key cellular mechanisms including RNA splicing, RNA modifications and post-transcriptional regulation. Although the factors of RNA regulation and the elements of RNA interactome are explored in the literature progressively, effective computational techniques, which can shed light on the diverse RNA landscape are still lacking. Therefore, we aim to elucidate the interactions between RNA transcripts and various factors of RNA interactome using the sequence information by using the exceptional predictive power of deep convolutional neural networks and vast amount of high-throughput sequencing data. Our novel method uses wide-range context information from RNA sequences, not limited to short RNA motifs. At present, we utilize CLIPSeq, HITS-CLIP, iCLIP and PAR-CLIP datasets from the doRiNA database in order to learn sequence features of RBP and RNP binding sites, beyond the limited length of RBP motifs. We are planning to further incorporate more RNA interactome resources such as RNA m6a-seq methylation sites and microRNA binding positions. This predictive method of the RNA landscape can be used to analyse mutation effects and their putative transcriptome-wide effects. 114 PP 7-09 Integrant roles of microRNAs and transcription factors in ovarian cancer transcriptional regulatory network Gov, E.1, Arga, K. Y.1 1 Marmara University, Bioengineering, Istanbul, Turkey Question: Transcriptional regulation of gene expression is often the primary process due to response of system alterations, wherein the information contained in a genome is converted and then eventually used to produce the proteins required for a given response. Methods: Here, we set out to reconstruct a transcriptional regulatory network of Homo sapiens consisting of experimentally verified interactions on miRNAs, TFs and their target genes. We have performed topological analyses to determine the transcriptional regulatory roles of miRNAs and TFs. Differentially expressed genes (DEGs) were identified by using two independent ovarian cancer associated transcriptome datasets (GSE7463 and GSE14407) from the Gene Expression Omnibus (GEO) database. Integration of differential expressed gene and miRNA data which was extracted from miR2disease database, on ovarian cancer was achieved to observe dynamic patterns of the disease specific gene expression. Results: As the regulators, the TFs HIF1A, TP53, EGR1, ETS1 and GATA3 as well as the miRNAs miR-30a-5p, miR-429, miR-206, miR-142-3p, and miR-30e-5p dominated the disease-specific sub-network regulating excessive number of DEGs. Overall results provided that positive and negative correlations occur in the active subnetwork and therefore mutual TFmiRNA regulation (interactive cooperation) and transcriptional regulations of target genes may occur through hierarchical mechanisms where miRNAs were the upstream regulators of TFs were the prominence strategies in gene regulatory network of ovarian cancer. In addition, multiple signals from miRNAs were integrated by TFs. Conclusions: Our results demonstrate new insights on TF and miRNA crosstalk, and here we present reciprocal interplay between ovarian cancer related miRNAs, TFs and their target genes. PP 7-10 Cigarette smoking increased expression of the G protein-coupled receptor 15 by change in CpG methylation Haase, T.1,2, Krause, J.1,2, Müller, C.1,2, Stenzig, J. 3,2, Röthemeier, C.1, Wild, P. S.4,5,6, Blankenberg, S.1,2, Zeller, T.1,2 University Heart Center Hamburg, Clinic for General and Interventional Cardiology, Hamburg, Germany German Center for Cardiovascular Research (DZHK e.V.), partner site Hamburg/Kiel/Lübeck, Hamburg, Germany 3 University Medical Center Hamburg-Eppendorf, Department of Experimental Pharmacology and Toxicology, Hamburg, Germany 4 University Medical Center of the Johannes Gutenberg-University Mainz, Center for Cardiology, Preventive Cardiology and Preventive Medicine, Mainz, Germany 5 University Medical Center of the Johannes Gutenberg-University Mainz, Center for Thrombosis and Hemostasis, Mainz, Germany 6 German Center for Cardiovascular Research (DZHK e.V.), partner site RhineMain, Mainz, Germany 1 2 OBJECTIVE: Cigarette smoking increases the risk for cardiovascular disease. Recent studies have focused on the effect of smoking habit on DNA methylation of target genes. Within the GPR15 gene, a novel cardiovascular risk candidate gene encoding the G-protein coupled receptor 15, the CpG locus cg19859270 has been identified to be hypomethylated in smokers, which was associated with increased GPR15 expression in white blood cells. The aim of this study was to analyze the longitudinal effect of smoking behavior on GPR15 expression in monocytes and peripheral mononuclear blood cells (PBMCs). METHODS: To investigate the effect of changes in smoking behavior and changes in expression over time, monocytic GPR15 mRNA expression was measured by qPCR in smokers (n = 179), ex-smokers (n = 369) and never-smokers (n = 480) from the population-based Gutenberg Health Study (GHS) at baseline and at the five-year follow-up visit. Linear mixed regression models were applied to test differential gene expression between baseline and follow-up visit. The methylation status of the GPR15 cg19859270 locus was determined in PBMCs from 262 subjects by bisulfite conversion and high resolution melting (HRM) and expression of mRNA of PBMCs was analyzed with the Affymetrix Exon ST1.0 GeneChip Array. Regression analysis was performed to estimate the association between smoking and GPR15 CpG methylation as well as gene expression. RESULTS: Monocytic GPR15 expression was significantly increased in smokers compared to ex- (p < 0.001) and neversmokers (p < 0.001). Ex-smokers had significantly elevated GPR15 expression levels compared to never-smokers up to ten years after quitting (p < 0.001). Smoking initiation between baseline and follow-up visit (17 subjects) led to an increase in GPR15 expression, whereas smoking cessation resulted in decreased GPR15 expression (42 subjects, p < 0.001). In PBMCs, the correlation between smoking and increased GPR15 gene expression (p < 0.0001) was associated with cg19859270 hypomethylation (p < 0.05). Likewise, smoking cessation was linked to cg19859270 methylation as well as reduced GPR15 expression (p < 0.05). Furthermore, higher cumulative smoking exposure correlated with lower GPR15 CpG site methylation and as well as higher GPR15 mRNA expression (p < 0.0001). CONCLUSION: This was the first longitudinal study showing that changing smoking behavior changed GPR15 expression. 115 Program / Abstracts This time- and dose-dependent effect was mediated by changes in DNA methylation. Hence, GPR15 poses an interesting new candidate gene for the link between smoking and cardiovascular disease. PP 7-11 Genetic analysis of spontaneous (non-toxic) liver fibrosis in a congenic mouse model Hall, R.1, Hochrath, K.1, Lammert, F.1, Grünhage, F.1 1 Saarland University Medical Center, Homburg, Germany Background: Mutations in the ABCB4 (ATP-binding casette, subfamily B, member 4) gene cause cholestatic liver diseases including progressive intrahepatic familial cholestasis (PFIC). Modifying genes of these diseases have yet to be identified systematically. In this study we used the Abcb4 (Mdr2) knockout (-/-) mouse model, in which the deficiency of the hepatobiliary phophatidylcholine floppase leads to chronic cholestasis, liver injury and fibrosis. As different mouse strains show varying fibrosis susceptibility, we applied a systematic approach to elucidate the genetic control of liver fibrosis in an experimental cross of ABCB4 deficient mice. Methods: The Abcb4 -/- knockout was crossed from the fibrosis-resistant FVB-Abcb4 -/- mice to the susceptible BALB/cJ strain by repeated backcrossing. To identify genetic modifiers that contribute to the fibrosis susceptibility linked to ABCB4 deficiency, we crossed these two congenic strains to generate an F2 intercross population. By quantitative trait locus (QTL) analysis differences in disease progression were mapped to polymorphic genetic regions across the whole genome. Single and two-dimensional QTL scans were applied to identify modifiers and pairwise gene interactions. Results: Compared to FVB-Abcb4 -/- mice, the BALB-Abcb4 -/- mice progress to higher fibrosis stages. The heterogenic F2 population shows marked phenotypic variation. Whereas single modifiers demonstrate minor effects, gene-gene interaction scans identified a significant interaction of two QTLs on chromosomes 4 and 17. Underlying these loci we identified the genes Abcg5, Abcg8 and sterol carrier protein 2 (Scp2) that are functionally related with hepatobiliary cholesterol homeostasis and resemble creedal modifer genes. Conclusions: The congenic Balb-Abcb4 knockout mouse allows the genomic exploration of a spontaneous, non-toxic disease model of a human gene defect. The experimental cross of the two genetic backgrounds with distinct fibrosis susceptibility enables the identification of Abcb4-dependent modifiers of cholestatic liver diseases. PP 7-12 ChIP-Seq pipeline for the identification of differential binding of the glucocorticoid receptor in high fat versus low fat diet in mouse livers Hawe, J.1, Quagliarini, F. 2, Uhlenhaut, H. 2, Heinig, M.1 1 2 Deutsches Forschungszentrum für Gesundheit und Umwelt, Helmholtz Zentrum München, Institute of Computational Biology, 85764 Neuherberg, Germany Deutsches Forschungszentrum für Gesundheit und Umwelt, Helmholtz Zentrum München, Institute for Diabetes and Obesity, 85748 Garching, Germany Background: Glucocorticoid receptor (GR) is a nuclear receptor and important transcription factor (TF) involved in regulating immune response, metabolism and other significant biological processes, often mediated through the binding of glucocorticoids (GC) to GR. Due to its role in metabolism, we studied the effects of high fat diet (HFD) as compared to normal/low fat diet (LFD) on the GR genomic binding in mice. Materials and methods: Liver samples were taken in a regular interval of 4 hours over 24 hours from mice exposed to the individual diets, first 5 days and then 12 weeks after start of the diet. A total of 48 samples, including replicates, were characterized by chromatin immunoprecipitation combined with sequencing (ChIP-Seq). We developed a specialized computational pipeline for the analysis of ChIP-Seq data, utilizing well established methods for quality control and the identification of TF bound chromatin regions, as well as a custom data normalization procedure and further downstream analysis methods. The normalization step included random down-sampling of the reads to a certain threshold, in order to account for variance between samples introduced during the sequencing process. To establish a set of well-defined, GR bound regions based on our data, we first identified potential peaks in the individual samples using standard peak calling methods and then built a universe of peaks reflecting possible GR binding sites by 1) retaining only peaks overlapping between replicates and 2) merging overlapping regions from different samples. This peak universe (17736 regions) then served as a basis for all downstream analysis. In addition, we identified and removed hidden confounding variables from the data via a factor analysis method, thereby getting rid of unwanted noise present in the data. The data adjusted for the identified confounding variables was then used for differential binding analysis. Results: We identified a total of 225 differential GR regions at a 10% false discovery rate. To further characterize the functional importance of those regions, we applied standard functional enrichment tools. Here we detected diverse mouse phenotype annotations, related to unhealthy liver and disturbed fat metabolism. The respective genes are currently subject to further experimental validation. In the future, this data set will also allow us to study the GR/GC circadian dynamics with respect to distinct diets in rodents. 116 PP 7-13 EXPERIMENTALLY-BASED MECHANISTIC MODELING TO STUDY T HELPER 17 CELL DIFFERENTIATION Intosalmi, J.1, Chan, Y. H.1, Rautio, S.1, Lähdesmäki, H.1 1 Aalto University, Department of Computer Science, Espoo, Finland Objectives: T helper 17 (Th17) cell differentiation is steered by extracellular cytokine signals that activate and control the lineage specific transcriptional program. Experimental studies provide us with an extensive amount of information about the Th17 lineage specific regulatory network but precise mechanistic understanding of the transcription factor dynamics is yet to be attained. Our main objective is to develop experimentally-based mechanistic modeling approaches that can be used to unravel the key molecular mechanisms steering the Th17 lineage specification and cell differentiation processes in general. Methods: To infer Th17 lineage specific regulatory mechanisms using time-course data, we combine mathematical modeling with advanced statistical techniques. We construct mathematical models in the form ordinary differential equations (ODEs) and link the models with experimental data by using well-defined statistical models. The ODE model design can be based on both known and hypothetical molecular interactions. In our computational implementation, we use state-of-the-art numerical and statistical methods, including population-based Markov chain Monte Carlo sampling and thermodynamic integration. Results: We apply our modeling approach to study the core network steering Th17 cell differentiation. We describe the core molecular dynamics by means of ODEs and calibrate the model parameters as well as the model structure in a data-driven manner using time-course RNA sequencing (RNA-seq) measurements. Within our modeling framework, the statistical properties of discrete read count RNA-seq data are taken into account by linking the ODE modeling with the observations through the negative binomial distribution. Our results show that we are capable of approving many known molecular interactions in the core Th17 network as well as predicting balancing effects that are caused by components of competing subsets of Th cells. Conclusions: By means of our modeling approach, it is possible to recover and predict core regulatory interactions that steer the Th17 lineage specification. Our results also illustrate how experimentally-based mathematical modeling enables knowledge and data driven model construction which takes the analysis of time-course data well beyond standard statistical analyses. PP 7-14 Uncovering potential therapeutic targets and prognostic biomarkers of ovarian tissue related diseases via systems biomedicine approach Kori, M.1, Gov, E.1, Arga, K. Y.1 1 Marmara University, Bioengineering, Istanbul, Turkey Question: Dysfunctions and disorders in the ovary may lead to various clinical or subclinical diseases, including ovarian cancer, ovarian endometriosis and polycystic ovarian syndrome (PCOS). Although the etiological relationship between ovarian cancer and endometriosis has been studied, the molecular mechanisms behind their connections are not yet in certain. Additionally in the light of the possibility of developing ovarian cancer is rising in women with PCOS, association among the diseases is feasible. The availability of high-throughput functional genomics (i.e., transcriptomics, proteomics, and metabolomics) data will expedite the understanding of the molecular mechanisms behind ovarian diseases. Methods: In the present study, we performed statistical methods for the analysis of transcriptomics data for ovarian cancer, ovarian endometriosis and PCOS, and integrated them with genome scale biological networks. Consequently, biomolecule (i.e: hub protein, transcription factor, miRNA) signatures at proteome, metabolism and transcription regulation levels were determined via integrative analyses, which might be feasible to uncover the novel biological mechanism insights behind the ovarian diseases. Results: In the study, it was determined (i) potential tendency of PCOS and ovarian endometriosis to tumorigenesis, (ii) molecules and pathways that related with cell cycle, apoptosis and MAPK signaling is the common indicator in formation and progression of all three diseases and (iii) significant association between PCOS and neurodegenerative diseases. Conclusions: To the best of our knowledge, this is the first report that illustrates the three ovarian diseases from systems biomedicine perspective. This study proposed signatures that could be considered as potential therapeutic targets together with the prognostic biomarkers in further experimental and clinical applications. 117 Program / Abstracts PP 7-15 Transcriptional profile of Aldara induced skin-reaction in humans Krause, L.1, Garzorz, N. 2, Lauffer, F. 2, Eyerich, S. 3, Theis, F. J.4,1, Eyerich, K. 2, Mueller, N. S.1 Helmholtz Center Munich, Institute of Computational Biology, Neuherberg, Germany Technical University, Dermatology and Allergy, Munich, Germany 3 Helmholtz Center and Technical University, Center of Allergy and Environment (ZAUM), Munich, Germany 4 Technical University, Mathematics, Munich, Germany 1 2 Imiquimod (Aldara) is used in mouse studies to mimic human psoriasis to test new therapeutic strategies and understand the underlying pathogenesis. Our aim is to understand what reaction Aldara induces in human skin. Over a time course of 30 days Aldara cream was regularly applied on 14 different patients. For comparison we used patients with different inflammatory skin diseases like psoriasis and allergic contact dermatitis (ACD). During the course of the experiment the patients were clinically monitored and several skin biopsies were taken for histological and gene expression analysis. Gene expression was measured using Agilent microarrays and modeled with linear mixed effect models to take into account inter-individual dependencies. Interestingly, the induced skin reactions rather show characteristics of eczema than hallmarks of psoriasis on the level of histopathology. Gene expression fold changes in Aldara treated skin show highest correlation with ACD skin (r = 0.78) and lower, yet significant, correlation with psoriasis skin (r = 0.57). Moreover, the overlap of significantly differentially expressed genes between Aldara treated skin and psoriasis is only 31% compared to 65% overlap between Aldara skin and ACD. In summary, Aldara induced skin reactions do not show the same histological and gene expression properties as seen in psoriasis. Whether some aspects of the disease might be reflected in the Aldara model is a question for further investigations. PP 7-16 A next generations sequencing toolbox with application to schizophrenia Hastreiter, M.1, Jeske, T.1, Hoser, J.1, Mewes, W.1, Küffner, R.1 1 Helmholtz Center Munich, IBIS, Neuherberg, Germany Motivation/question: Success of large-scale data analysis depends on sophisticated bioinformatic support to process, integrate, analyse, and interpret Big Data volumes. An example for such Big Data is Next Generation Sequencing (NGS) data. Analysis of NGS data requires chaining of various tools with complex input and output formats. This makes reliable and efficient data handling difficult for non-expert users. Methods: In order to cope with the increased throughput of massive data generating experiments we are relying on KNIME, the Konstanz information miner. This system allows to create structured and reusable processing pipelines for all kinds of analyses that are easy to use and can be shared among the community. Results: Therefore we have developed a comprehensive KNIME toolkit including nodes for typical steps like read preprocessing, read mapping, variant calling, detection of differential expression and annotation. Complementary to existing nodes, our toolbox now facilitates the assembly of basic building blocks into a wide range of customized NGS analysis workflows. We also developed a High-Throughput Executor (HTE) for KNIME, in order to process complex datasets with minimal intervention. In schizophrenia, so called “de novo” mutations found in affected children but not in their healthy parents are likely involved. Thus, the genotype could starting point for both therapy and diagnosis. Here we examined 36 well characterized family trios and several affected single individuals resulting in 278 cases and 72 parents. Our approach combined exome, genome and metabolome techniques and the detection of associations between large datasets and clinical phenotypes We present an automated, ready to use pipeline for the identification and characterization of variants. The analysis workflow for processing NGS data and the detection of de novo and loss-offunction (LOF) variants is depicted in the figure. Conclusion: Compiling the necessary tools into standardized graphical workflows improves transparency and adaptability of analysis pipelines and can therefore stimulate collaboration between computational and wet lab biologists. The publication of scientific results together with the generating workflows furthermore has the potential to improve representation, reproducibility, and dissemination of findings substantially. We demonstrated the advantages of our workflows based on a systems medicine approach to discover schizophrenia disease mutations. 118 PP 7-17 The properties of Ewing Sarcoma driver cells - widen the view beyond the master fusion gene Mallela, N. V.1, Hotfilder, M. 2, Seggewiss, J. 3, Jakalski, M. T.1, Dirksen, U. 2, Korsching, E.1 Institute of Bioinformatics, University of Muenster, Muenster, Germany Pediatric Hematology and Oncology, University Children’s Hospital, Muenster, Germany 3 Institute of Human Genetics, University Children’s Hospital, Muenster, Germany 1 2 Main Objectives: Ewing’s Sarcoma (ES) belongs to the group of bone cancers defined up to now by the existence of one fusion gene. This oncogenic fusion consists of EWSR1 and members of the ETS gene family. In this study we use the model cell line CADO-ES1 (EWSR1-ERG) to characterize the tumor biology of ES stem cell like driver cells subpopulation versus the main population of tumor forming cells. We aim to compare these two subtypes to identify specific characteristics beyond the fusion gene concept which does not form a sufficient model to explain Ewing sarcoma biology. We also aim to compare these results to the properties of mesenchymal stem cells and fibroblast cells to gain finally a better understanding of the interplay of the ES constituting cell subpopulations and possible therapeutic interventions. Materials and Methods: The CADO-ES1 cultured cells were FACS sorted into SP (side population) and nonSP (non proliferative) and extensively characterized. Transcriptomic sequencing and whole genome sequencing of CADO-ES1 of all cell (sub)types was done. Established bioinformatics tools (GATK, Tophat, edgeR, defuse, STAR) were used to conduct various types of analysis. Inhouse tools were also developed to facilitate the analysis of colorspace reads with standard bioinformatics tools. Results: In addition to the EWSR1-ERG fusion, we have also found the NAIP-OCLN fusion gene in the SP cells. Although this has not yet been discussed in the field of ES, this fusion gene has been reported in the small cell lung cancer (Iwakawa et al., 2013). Looking at the DNA repair mechanisms, the expression of the e.g. gene PARP1 has remained the same in both SP and nonSP cells. We have also observed that PTEN has almost no expression in both the cell types and there is no differential expression of TP53. Apart from the above observations, we have also done a variant analysis of the ES transcriptome and genome and have so far observed a SNP in CHN2 that is presumed to have a disruptive impact at the protein level. We also observed a significant down regulation of its expression in SP cells. With a family of further candidates the expression results are coherent with the observations by Hu-Lieskovan et al. 2005. Conclusion: The whole amount of observations up to now underline that the fine structure of ES biology and the search for cell populations with complementing sub tasks are necessary to get closer to an explanatory concept of the systemic stability and the crucial power of ES. PP 7-18 Genome-wide interplay of DNA methylation with genetic variants on the human metabolism Molnos S.1 1 Helmholtz Zentrum München , AME, Munich, Germany In the past years, through genome-wide association studies (GWAS) hundreds of genetic variants were found to be associated with metabolites that are involved in relevant biological processes. However, so far none study have systematically, on a large scale investigated an interplay between genetic and epigenetic changes such as DNA methylation. In this study we calculate a genome-wide analysis of the interaction between genetic variants and DNA methylation sites on metabolites in order to enhance the understanding of important biological processes. Metabolite data were measured by three different metabolomics platforms using a kit-based-targeted quantitative FIA-MS/ MS method (‘Biocrates’), non-targeted, semi-quantitative LC-MS/MS and GC-MS methods (‘Metabolon’), and 1H NMR measurements in order to derive lipid-related parameters (‘LipoFIT’). To reduce the number of potential models, CpGs and SNPs with a superthreshold association value to one of the metabolites (p-value < 0.05) were filtered. Finally, interaction analysis was conducted using the sample correlation |r| as test statistic. P-values were subjected to multiple testing corrections using the Bonferroni correction. In the lipofit platform most associations were found for the lipoprotein fractions small and medium HDL and large LDL. In the biocrates platform most associations were found for amino acids, whereas in the metabolon platform most associations were found for the metabolite butyrylcarnitine. Furthermore, interacting SNPs and CpGs can be on different chromosomes. Interestingly, for the metabolon platform, we found strongest associations with the metabolite butyrylcarnitine and the corresponding SNPs and CpGs were only close to each other on the same chromosome in contrast to associations with metabolites measured by the other platforms. 119 Program / Abstracts PP 7-19 Reconstruction of the FANCA interactome and discovery of new putative oncogenes in head and neck squamous cell carcinoma. Pitea, A.1, Sass, S. 2, Theis, F. J. 2,3, Zitzelsberger, H.4, Unger, K.4, Müller, N. S. 2 Helmholtz Zentrum München Research Center for Environmental Health (GmbH), Institute of Computational Biology/Research Unit Radiation Cytogenetics, Neuherberg, Germany Helmholtz Zentrum München Research Center for Environmental Health (GmbH), Institute of Computational Biology, Neuherberg, Germany 3 TUM Technische Universität München, Mathematics, Garching, Germany 4 Helmholtz Zentrum München Research Center for Environmental Health (GmbH), Research Unit Radiation Cytogenetics, Neuherberg, Germany 1 2 Head and neck squamous cell carcinomas (HNSCC) are regularly treated with radiotherapy alone or concomitant with chemo- or immunotherapy. Patients suffering from HNSCC present an overall 5-year survival rate of approximately only 50% which is likely to be attributed to radioresistance. A previous study on patients that were treated with radiotherapy alone identified a genomic copy number gain of chromosome 16q24.3, containing the DNA repair gene FANCA, which was associated with unfavourable outcome in radiation-treated patients. With this as a starting point, we set up to validate FANCA at multiple molecular levels in an independent cohort and to build an integrated framework. This would allow us to understand the FANCA associated mechanisms of radioresistance and to discover new putative HNSCC specific oncogenes as potential therapeutic targets. To study the molecular profile of radiotherapy treated patients on a multi-omics level, we used the HNSCC dataset available on The Cancer Genome Atlas (TCGA) portal. Copy number variation and clinical data allowed us to validate the copy number gain and, also, to characterize our subset at the genome level. We want to link clinical data with copy number changes, as well as with mRNA and miRNA expression levels of the investigated HNSCC cohort on genome-wide level. To integrate the three omics data, we have first designed a data-driven method, called miRlastic, for the identification of mRNAs targeted by microRNAs through a penalized elastic net model coupled to prior knowledge of target predictions. Functional annotation of the multi-omics regulation network was achieved by a scoring approach of the local neighbourhoods in the network. The next step lies in cross-linking the mRNA and miRNA level with genomic copy number data. The overall aim comprises the challenge of validating and identifying novel putative oncogenes involved in the regulatory network of HNSCC and genes that affect the patient outcome. PP 7-20 Modelling the crosstalk between DNA methylation and histone modifications in cancer Przybilla, J.1, Rohlf, T.1, Loeffler, M.1,2, Galle, J.1 1 2 Leipzig University, Interdisciplinary Center for Bioinformatics, Leipzig, Germany University Leipzig, Institute for Medical Informatics, Statistics and Epidemiology , Leipzig, Germany Aberrant DNA methylation is a characteristic feature of various types of cancer and in several cases has been used to classify tumor subtypes with different clinical outcome. It was shown that promoters associated with nucleosomes that carry trimethylation of lysine 27 at histone 3 (H3K27me3) are prone to become methylated in cancer. Tri-methylation of lysine 4 at histone H3 (H3K4me3) on the other hand is known to protect CpG-rich promoters from getting DNA methylated. Here we ask how crosstalk between different epigenetic regulators generates experimentally observed DNA methylation pattern. We introduce a computational model framework that allows simulating the transcriptome and the epigenome of stem cells during homeostasis and cancer development. Each cell in the model contains an artificial genome. The artificial genome encodes a transcription factor network that, together with histone modifications and DNA methylation, controls gene expression, which itself feeds back on the epigenome. In simulations of our model we analyze the regulatory states of more than hundred cells, thus bridging the molecular-, cellular-, and population scale of description. We demonstrate that increased proliferation and aberrant activity of epigenetic modifiers can affect the stability of histone modifications and consequently the DNA methylation pattern. We find that hyper-methylation of CpGs following accelerated proliferation occurs frequently at regions that are associated with H3K27me3 modified chromatin, whereas hypo-methylation occurs in regions associated with stable H3K4me3. Our model suggests that the specific local combination of histone modifications determines the possible methylation changes of the associated DNA. We validate this prediction by re-analyzing data of the TCGA consortium on DNA methylation in colorectal carcinoma samples. In summary, our computational model allows for the first time to simulate how modified activity of DNA methylation and histone modifiers in concert might determine the development of specific DNA methylation pattern in cancer. 120 PP 7-21 Importance of rare gene copy number alterations for personalized tumor characterization Seifert, M.1, Friedrich, B. 2, Beyer, A. 3 Institute for Medical Informatics and Biometry (IMB), Faculty of Medicine Carl Gustav Carus, TU Dresden, Dresden, Germany Institute of Molecular Systems Biology, Zurich, Switzerland 3 Cellular Networks and Systems Biology, CECAD, University of Cologne, Cologne, Germany 1 2 Copy number alterations (CNAs) of large genomic regions are frequent in many tumor types, but only few of them are assumed to be relevant for the cancerous phenotype. It has proven exceedingly difficult to ascertain rare mutations that might have strong effects in individual patients. New systems biological approaches are urgently needed to overcome this. Here, we show that a genome-wide transcriptional regulatory network inferred from gene expression and gene copy number data of 768 human cancer cell lines can be used to quantify the impact of individual patient-specific gene CNAs on cancerspecific survival signatures using a specifically designed network impact propagation algorithm. The model was highly predictive for gene expression in 4,548 clinical samples originating from 13 different tissues. Focused analysis of tumors from six tissues revealed that in an individual patient a combination of up to 100 gene CNAs directly or indirectly affect the expression of clinically relevant survival signature genes. Importantly, rare patient-specific gene CNAs (less than 1% in a given cohort) often have stronger effects on signature genes than frequent gene CNAs. Subsequent integration with genomic data suggests that frequency variation among high-impact genes is mainly driven by gene location rather than gene function. Survival analyses on independent tumor cohorts revealed tumor-type specific trends indicating that rare gene CNAs can be as important as frequent gene CNAs for the prediction of patient survival. Our framework contributes to the individualized quantification of cancer risk, along with determining individual key risk factors and their downstream targets. PP 7-22 Microarrays in large-scale phenotyping at the GMC Söllner, J.1, Fuchs, C.1, Theis, F. J.1 1 Helmholtz Zentrum München, ICB, Neuherberg, Germany At the German Mouse Clinic (GMC) mouse knockouts (KO) are not only phenotyped with respect to e.g. clinical, behavioural or neurological parameters but also their gene expression is analysed. Thus, a dataset of 28 Illumina BeadChips is available comprising 219 samples covering nine organs and ten genetic backgrounds. In most studies, one is interested in determining differentially expressed genes between mutant and wildtype mice of one specific KO line. Here, in contrast, we aim at assessing whether there are strain- or tissue-specific patterns in the transcriptome. In a largescale approach we investigate all available datasets simultaneously. The findings would facilitate future studies across mouse strains and tissues. For example, one could then compare transcriptomics data of two KO lines with different genetic backgrounds. Knowing which genes are differentially expressed merely due to strain-specific patterns helps to narrow down the overall gene list to those relevant for further analysis. PP 7-24 ModuleDiscoverer: a novel approach to the discovery of disease modules refining molecular characterization of diet induced rodent models of fatty liver. Tokarski, C.1,2,3, Guthke, R.1, Schuster, S. 2, Dahmen, U. 3, Vlaic, S.1,2 Hans-Knöll-Institute (HKI) Jena, AG Systemsbiology and Bioinformatics, Jena, Germany Friedrich-Schiller-University, Department of Bioinformatics, Jena, Germany 3 University Hospital Jena, Friedrich-Schiller-University, General, Visceral and Vascular Surgery, Jena, Germany 1 2 Diet induced rodent models of fatty liver disease are essential in numerous areas of research where the availability of human tissue is limited due to ethical as well as practical reasons. Therefore, model specific characterization of molecules, mechanisms and processes involved in the induction and maintenance of steatosis is important for the translation of experimental results to human pathophysiology of the fatty liver. In the science of networks, recent publications highlight the advantages of disease module discovery to increase our understanding of diseases on the mechanistic as well as the systems level. Thus, identification of diet specific disease modules has the potential to enhance existing characterization of rodent models of fatty liver disease (FLD). Given published microarray data of four high-fat (HF) diet induced rodent models of the fatty liver we identify diet specific differentially expressed genes resembling the molecular alterations that can be attributed to the source of fat in the diet, namely lard (HF-L), olive oil (HF-O), coconut fat (HF-C) and fish oil (HF-F). We then introduce the ModuleDiscoverer algorithm, a new approach for the ab initio identification of disease modules using protein-protein interaction networks in conjunction with differentially expressed genes. Our results show that the identified disease modules for HF-O, HF-F, HF-C 121 Program / Abstracts are significantly enriched with genes associated to diseases relevant to fatty liver. Structural analysis of the derived dietspecific disease modules revealed key-regulatory genes with central roles in the modules. Among already well known FLDassociated genes, our findings enclose new key-genes potentially involved in the induction of FLD in the respective rodent models. We present ModuleDiscoverer, an algorithm suitable for the ab initio identification of disease modules in protein-protein interaction networks. Its heuristic approach to the detection of communities allows processing of real-world biological networks at low computational costs. This makes ModuleDiscoverer favorable over current computationally heavy approaches that, for example, rely on the enumeration of all maximum cliques. Application to experimental data reveals new insides to mechanisms involved in the induction of FLD. While all diets are known to induce FLD in rodents, on the molecular level the HF-O diet shows the widest spectrum of human FLD associated pathophysiological characteristics, followed by HF-F and HF-C. PP 7-25 Differential gene expression in atrial fibrillation patients von den Driesch, L.1, Krause, J. 2,3, Müller, C. 2,3, Rzayeva, N. 2,3, Schillert, A.4, Schnabel, R. 2,3, Zeller, T. 2,3, Heinig, M.1 Helmholtz Zentrum München, Institute of Computational Biology, Neuherberg, Germany University Heart Center Hamburg, Clinic for General and Interventional Cardiology, Hamburg, Germany 3 German Center for Cardiovascular Research (DZHK e.V.), Hamburg, Germany 4 Universität zu Lübeck, Institut für Medizinische Biometrie und Statistik, Lübeck, Germany 1 2 Objective: Atrial fibrillation (AF) is the most common arrhythmia in the general population with high comorbidity, mortality and significant public health impact. So far, the molecular mechanisms underlying AF are largely unknown. The first step of investigating candidate genes, which associate with AF, was the analysis of the differential gene expression of population-based cohort with 83 individuals including 26 AF cases. Materials and methods: We performed transcriptome analysis in right atrial appendage tissue samples of 84 patients, who had a bypass surgery. Among those 26 AF cases, some were diagnosed with prevalent AF, while others developed incident AF after surgery. Gene expression values of mRNA and lincRNAs were measured on an Affymetrix Gene 2.0 ST chip. Visual quality control was performed using principle components analysis, batch effects were removed using combat and potential hidden confounders were detected using PEER. To focus on the AF related gene expression changes, the ongoing analysis introduces three groups of cardiovascular risk factors and biomarkers into a linear model to adjust for the influence of other confounding factors. These groups consist of a simple correction by age and sex, medical conditions (e.g. Myocardial Infarction) and molecular phenotypes (e.g. Nt-proBNP). Furthermore, to identify the underlying regulatory mechanisms, the differentially expressed genes will be integrated with predicted transcription factor binding affinities to highlight the transcription factors related to AF. Conclusion: We aim to verify already published and identify novel candidate genes, which correlate both with prevalent and incident AF cases. After confirming those differentially expressed genes by quantitative PCR, the next step will be a molecular characterization by using different in vitro approaches including siRNA-mediated silencing or targeted overexpression of the identified transcription factors, to confirm their direct target genes. PP 7-26 Integrative analysis of exome, transcriptome and cellular bioenergetics profiles in patients with rare mitochondrial diseases Yepez, V.1, Kremmer, L.1, Bader, D.1, Prokisch, H.1, Gagneur, J.1 TUM, Department for Bioinformatics, Garching, Germany 1 Multi-omics studies that include analyzing the exome, transcriptome, proteome and phenome in an integrated manner are gaining popularity because they can provide deeper insights into biological systems. As such, we investigate certain phenotypic traits of fibroblast cell lines, originated from patients with rare mitochondrial diseases. Because the predominant physiological function of mitochondria is the generation of ATP by oxidative phosphorylation (OXPHOS) during cellular respiration, we are interested in measuring traits related to that biological process. So far, we have oxygen consumption rate (OCR) and extracellular acidity rate (ECAR), that reflect OXPHOS and glycolytic metabolism, respectively. Taken at different experimental conditions, OCR and ECAR can give insight into specific bioenergetics like basal and maximal respiration and ATP production. We then compare the results for each patient with their transcriptome and find genes related to each possible dysfunction. The long term scope of the project is to join these findings with ones obtained from exome and proteome analyses and do a full multi omics study. 122 INDEX A Abdukerim A. 111 Abou-Jaoudé W. 67 Adamson A. 38, 57, 60 Adlung L. 44, 67 Adlung L. 68 Admon Y. 46 Adornetto G. 103 Ahnert P. 109 Ahrends R. 100 Alessandro L. 59 Aliluev A. 57 Alkan O. 45 Allgöwer F. 62 Alon U. 49 Alpert A. 100 Andersen M. K. 52 Anderson A. 32 Andriasyan V. 54 Andrieux G. 68 Angerer P. 55 Ankers J. 60 Appel L. M. 101 Arga K. 112 Arga K. Y. 79, 115, 117 Arvaniti E. 56 Asara J. 104 Asbeh N. 46 Ashall L. 38 Avsec Z. 111 Awais R. 60 Aydin B. 112 B Bacall F. 90 Bader D. 122 Bagnall J. 35 Baier V. 103 Bajikar S. S. 58 Bakermans O. 93 Bakker B. 46 Balbus J. 84 Ballnus B. 101 Balsa-Canto E. 62 Banga J. 62 Bao J. 67 Barberis M. 111 Baretton G. B. 114 Bartenschlager R. 39 Bathen T. F. 52 Baum K. 39, 68 Bazil J. 93 Beard D. 93 Bruggeman F. 82 Dalmasso G. 88 Beaudette P. 39 Brühlmann D. 41 Damm G. 48, 39 Beckers J. 57 Brümmendorf T. 96 Daniels D. 38 Becker T. 114 Brummer, T. 28 Davies D. 47 Belderbos M. E. 46 Brusch L. 35, 102, 89, 114 de Back W. 42, 89 Belicova L. 53 Buard V. 91 Deharde D. 48 Bell E. 36 Buchholz V. 58 de Jonge N. 70, 46 Benary U. 69 Bueb J.- L. 85 de Landtsheer S. 74 Bergey F. 113 Buettner F. 58 del Mistro G. 74 Bergheim I. 31 Buettner M. 58 Depner S. 44, 79 Berndt N. 83 Bulik S. 83, 102 Deshpande S. 108 Berndt N. 82 Bullinger E. 77 Deutsch A. 35, 89 Bertilsson H. 52 Burkhart J. M. 99 De Vos W. 36 Beyer A. 121 Busch D. 58 Diaz-Zuccarini V. 97 Beyer H. 69 Busch H. 80 Diego J. 29 Bielow C. 82 Busch H. 68, 67 Diken M. 101 Billing U. 56, 77 Buske P. 42 Dirksen U. 119 Binder M. 39, 78 Büttner F. 55 Dittmar G. 39 Bitzer M. 94 Büttner M.1 55, 57 Dittrich A. 56 Dittrich A. 77 Björkegren J. 34 Björkegren J. L. 110 C Divine M. 89 Björkhem I. 87 Campillos M. 103 Dooley S. 91 Blankenberg S. 115 Cappell S. 47 Dörffler V.4 76 Blank L. 103 Carpenter A. E. 47 Downton P. 35, 38 Blasi T. 47 Catena R.1 54, 65 Drabløs F. 52 Blüthgen N. 28 Cedersund G. 44, 47 Dräger A.1 81 Boddington C. 35, 38, 57 Celliere G. 43, 88 Drasdo D. 43, 54, 88, 92 Bode J. G.4 39 Cerveira J. 47 Drayman N. 49 Bodenmiller B. 30, 54, 65, Chakraborty S. 39, 68 Drummond D. 45 Chan P. 91 Dubovik T. 46 66, 61, 64, 81 Boegel S. 105 Chan Y. H. 117 Boekholdt M. 93 Chaouiya C. 67 E Boerries M.4 80, 5 67 Chara O. 35 Ebrahim A. 81 Bohl S. 80 Chaudhuri T. 108 E. de Bont S. J. 46 Böhm M. E. 39, 44, 79 Chis O.- T. 62 Egners A. 82 Boissier N. 43, 88 Christen H. 77 Ehlting C. 39 Boley D. 85 Chung M. 47 Ehninger G. 114 Borgos S. E. 52 Claassen M. 56 Eils R. 60 Börner K. 59 Clarke K. 71 Eisfeld A. 40 Borries M. 68 Clausznitzer D. 78 Eissing T. 104 Borth N. 83 Colomé Tatché M. 46 Enard W. 38 Böttcher A. 57 Coman C. 100 Engesser R. 69 Botvinick E. 51 Cordes H. 103 England H. 113 Boyd J. 35, 38, 60 Corne T. 36 Engström M. 47 Bradley A. 46 Cramer T. 82 Eraslan G. 114 Braun, S. 28, 59, 72, 80 Cvitanović T. 40 Erdem M. 82 Erdmann T. 83 Brehme M. 96 Breuhahn K. 54 D Erhart W. 102 Bridge L. 60 Dahl A. 114 Eyerich K. 118 Brieu R. 95 Dahmen U. 122 Eyerich S. 118 Brignall R. 35 d’Alché-Buc F. 91 Broly H. 41 D’Alessandro L. A. 92, 80 Brosch M. 114, 102 Dallmann A. 104 123 Program / Abstracts F Guthke R. 80, 122 Faryna M. 101 Holzhütter H. 95 Kanev K. 66 Holzhütter H. - G. 80, 82, 83 Kania R. 90 Fehling-Kaschek M. 70 H Hopf T. 92 Kankeu C. 71 Ferronika P. 46 Haanstra J. 82 Hoppe A. 80 Kaoma T. 85 Feuerhake F. 95, 98 Haase T. 115 Horan S. 51 Karlsson M. 47 Fiedler A. 70 Haghighi E. 113 Horger M. 94 Kar S. 67 Filby A. 47 Haghverdi L. 55, 58 Horn M. 59 Kaschek D. 70, 72, 76, 77, 107 Filiou M. 104 Hahlbrock J. 105 Hoser J. 118 Kastenmüller G. 61 Findeisen R. 77 Hall R. 116 Hoshino D. 78 Katsura H. 40 Florez A. 36 Halsema N. 46 Hotfilder M. 119 Kauczor H.- U. 54 Floßdorf M. 58, 101 Hamacher-Brady A. 52, 88 Howard M. 34 Kawakami E. 40 Foijer F. 46 Hammad S. 91 Huang X. 39, 59, 72, 80 Kawaoka Y. 40 Forestier G. 95 Hampe J. 53, 102, 114 Huber H. 70, 71 Kawata K. 78 Foroughi Asl H. 110 Hanscho M. 83 Hucho T. 63 Kazemier H. G. 46 Forsgren M. 47 Harder N. 59 Hug S. 71, 101 Kegel V. 48 Förster K. 108 Harper C. 38, 60 Huppelschoten S. 72, 76 Keller S. 71 Franke A. 44 Hasenauer J. 4 3, 60, 62, 63, Hutter S. 41 Kempa S. 82 70, 71 Frankish J. 78 Kessler T. 43,110 Frensing T. 28 Hasenauer J. 101 I Khatri P. 100 Freyer N. 48 Hasenclever D. 110 Iber D. 51 Kheifetz Y. 107 Friebel A. 88 Hassan R. 49, 109 Ince I. 104 Kierzek A. M. 33 Friedrich B. 121 Hassel J. 101 Ingraham J. 92 King Z. A. 81 Fritsche-Guenther, R. 28 Hass H. 79 Intosalmi J. 117 Kitano H. 40 Fröhlich F. 43 Hastreiter M. 118 Ipenberg I. 39 Kittner M. 113 Fröhlich F. 60, 62 Hatano A. 78 Irmler M. 57 Klarner H. 73 Fuchs C. 58, 83, 121 Hatzikirou H. 95 Israel A. 63 Klemm S. 28 Fujii M. 78 Hatz K. 101 Ito Y. 78 Klingmueller U.2 92 Fuji K. 40 Hausser A. 70 Itzkovitz S. 36 Klingmüller U. 39, 44, 67, 68, Hawe J. 116 Hecht A. 69 J Klipp E. 44 Gage M. 97 Heemskerk J. W. M. 99 Jackson D. 57, 60 Kneissl J. 71 Gagneur J. 111, 122 Heinäniemi M. 85 Jackson H. 54, 65 Koeppl H. 61 Galleguillos S. 83 Heinig M. 113, 116, 122 Jagiella N. 92 Kofahl B. 39, 68, 69 Galle J. 42, 120 Heinonen M. 91 Jaillet C. 91 Kok K. 46 Galliani S. 89, 97 Heldt F. S. 28 Jaimovich A. 47 Kondofersky I. 80, 83 Garzorz N. 118 Hellmann I. 38 Jakalski M. T. 119 König M. 95 Gebhardt R. 84, 86, 114 Hempel G. 104 Janes K. A. 58 Korf U. 70 Gebhardt R. 87 Hengstler J. 4 3, 49, 88, 91, Janssens S. 70 Kori M. 117 Jarsch M. 44 Korsching E. 119 92, 109 Georgi F. 54 124 72, 76, 79, 80, 92 G Georg J. 75 Hennig H. 47 Jaster R. 76 Koschmieder S. 96 Ghallab A. 43, 49, 109 Herberg M. 42 Jeneson J. 93 Koshkaryev A. 45 Glauche I. 42, 55 Hergenröder R. 91 Jeneson T. 94 Kramer B. A. 44 Goble C. 90 Herrmann F. 76 Jeruc J. 87 Krantz M. 73 Golebiewski M.1 90, 105, 106 Herr, R. 28 Jeske T. 118 Kranz L. M. 101 González-Vallinas M. 54 Hess W. 75 John E. 85 Krause J. 115, 122 Goria T. 40 H. Groen J. M. 46 Jones N. 60 Krause L. 118 Görlitz L. 101 Hilgendorff A. 108 Joshi T. 97 Krebs O. 90 Gov E. 115, 117 Hiller T. 48 Jünger A. 72 Kremmer L. 122 Grabe N. 54 Hiltermann T. J. N. 46 Jüngst C. 102 Kretschmer L. 58 Graef P. 58 Hinz J. 100 Juvan P. 87 Kretzschmar G. 102 Grandclaudon M. 67 Ho A. D. 44 Green V. 50 Hochrath K. 116 K Kreutz C. 79 Gretz N. 80 Hoehme S. 88, 92 Kaderali L. 76,78, 114 Kreutz C. 39 Grimm D. 59 Hoekstra-Wakker K. 46 Kakuda H. Kreuz M. 59 Grünhage F. 116 Höfer T. 36, 58, 101 Kubota H. 78 Kubica K. 84 Guillouzo A. 107 Höfer T. 67 Kalaidzidis Y. 53 Kubot H. 41 Guipaud O. 91 Hoffmann A. 30 Kaleta C. 85 Kuchen E. 36 Gunawan R. 41 Hofmann U. 109 Kalkan T. 42 Kuepfer L. 103, 109 Guryev V. 46 Hollmann S. 105, 106 Kallenberger S. 60 Kuepfer S. 109 Kreutz C. 87 Küffner R. 118 Mansour A. 100 N R Küffner R. 65 Marchesini G. 95 Nahlik T. 53 Raeth S. 70 Kulms D. 74 Marin Zapata P. A. 52 Navarro P. 105 Raghuraman S. 65 Kumar S. 61 Marioni J. 30 Nederveen A. 93 Rand D. 31 Kunida K. 78 Markowetz F. 37 Nguyen Q. 90 Rapsomaniki M. A. 64 Küpfer L. 114 Marks D. 92 Nicot N. 85 Rateitschak K. 76 Kupke S. Y. 28 Marks D.S. 28 Nikolaew A. 105 Raue A. 44, 45, 79 Kuritz K. 62 Marr C. 55 Nikos V. 85 Rautio S. 117 Kuroda S. 41, 78 Martin V. 70 Niranjan V. 108 Rees P. 47 Kurup A. 51 Mastrobuoni G. 82 N. K. N 108 Regierer B. 105, 106 Matsuoka Y. 40 N. Mattheij J. A. 99 Reichl U. 28 L Matz-Soja M. 84, 86 N. Müller S. 80 Reif R. 49 Laeremans T. 70 Matz-Soja M. 87, 114 Nordgård A. 52 Reinz E. 70 Lähdesmäki H. 117 Mazurov V.2 35 Normand R. 46 Rennert C. 84, 86 Lahrmann B. 54 McDonagh C. 45 Nouailles-Kursar G. 109 Reuss M. 89, 94, 97 Lam C.1 38 Meichsner J. 59 Nussbaumer M. 104 Rey M. 90 Lammert F. 102, 116 Menck K. 100 Nyman E. 44 Rise K. 52 Landgraf R. 104 Menni C. 61 Lang C. 51 Merino Tejero E. 82 O Robijns J. 36 Lange B. 43 Merkle R. 44, 79 Offringa R. 101 Rodrigo Albors A. 35 Lansdorp P. M. 46 Merkt B. 76 Ohno S. 78 Rodriguez-Gonzalez A. 44 Lapin A. 94 Mertes C. 111 Olesch J. 54 Rodriguez Martinez M. 61,64 Lattermann S. 44, 67 Mewes W. 118 Oppelt A. 72, 76 Roeder I. 42, 55 Lauffer F. 118 Meyer-Hermann M. 98 Oprea T. 32 Roelli P. 66 Lehmann, N. 28 Meyer K. 102 Ostrenko O. 102 Roesch K. 110 Lehmann W. D.1 44 Meyer M. 104 Othman A. 91 Rohlf T. 120 Lehmann W. D.2 79 Meyer R. 80 Owen S. 90 Rohr K. 59 Lehrach H. 43 Meyer T. 47 Leonhardt H. 38 Michos O. 51 P Ross E. 37 Lerman J. A. 81 Miller P. C. 81 Palsson B. O. 81 Rost F. 35 Lewis N. E. 81 Milliat F. 91 Papoutsopoulou S. 113 Röthemeier C. 115 Lickert H. 57 Mintser O. 75 Parekh S. 38 Rother K. 42 Ligon T. 62 Moeller K. 63 Passante E. 70, 71 Rowe W. 35, 38 Li J. 74 Molenberghs F. 36 Paszek P.1 3 5, 38, 57,60, 64, Rozman D. 40, 87 Liu X. 103 Möller P. 85 Loeffler M. 4 2, 59, 107, 109, Molnos S. 119 Patterson J. O. 47 Rudolph N. 77 Montazeri M. 96 Peckys D. 70 Rummel H. 56, 77 Loos C. 63 Monteiro P. T. 67 Pelkmans L. 50 Ryan S. 38 Lopez Alfonso J. C. 95 Morales-Navarrete H. 53 Perse M. 87 Ryan S. 60 Lorbek G. 87 Morbidelli M. 41 Pfau T. 85 Rychtarikova R. 53 Lotz J. 54 Morkel, M. 28 Pichardo-Almarza C. 97 Rye M. 52 Lübberstedt M. 48 Moškon M. 40 Pineda-Torra I. 97 Ryl T. 36 Luber B. 71 Mraz M. 40 Pires Pacheco M. 85 Rzayeva N. 122 Lucarelli P. 74 Muckenthaler M. U. 80 Pitea A. 120 S Lu J. S. 81 Muders M. 114 Poelwijk F. 92 Saber A. 46 Lundberg P. 47 Mueller N. S. 114, 118 Politi A. 68 Saez-Rodriguez J. 66 Lundengård K. 47 Mueller-Roeber B. 105 Pollak N. 62 Sahay A. 28 Lun X. 61, 64 Mueller W. 90 Porubsky D. 46 Sahin N. 105, 111 Lutter D. 63 Müller B. 54 Poschke I. 101 Sakabe S. 40 Müller C.2 101, 115, 122 Prasanna A. 108 Salopiata F. 44, 79 M Müller F. 62 Preusse M. 114 Sander C. 29, 92 Maccarrone G. 104 Müller N. S. 120 Prokisch H. 111,122 Sano T. 78 Mader W. 75 Müller S. 80 Przybilla J. 42, 120 Sass S. 57, 108, 120 Maemura T. 40 Muramoto Y. 40 Maiwald T. 39 Murphy M. 104 Q Satoshi F. 40 Malenka M. 75 Murphy R. 29 Quagliarini F. 116 Sauter T. 74, 85 120 Robichon K. 39 Rosenblatt M. 77 113 Ruckerbauer D. E. 83 Sato M. 78 Malkusch S. 76 Schaadt N. 95, 98 Mallela N. V. 119 Schafmayer C. 102, 114 Mansmann U. 74 Schaper F. 56, 77 125 Program / Abstracts Schapiro D. 54, 65, 66 Spierings D. C. J. 46 V Wolf J. 39, 68, 69 Scheidereit C. 39 Spiller D. 35, 38, 60 Vallar L. 85 Wolkenhauer O. 76 Schelker M. 44, 79 Springer M. 92 van den Berg A. 46 Wolstencroft K. 90 Schenk A. 109 Stanford N. 90 van den Bos H. 46 Wrona A. 84 Schild H. 105 Starosvetsky E. 46, 100 Van der Hoeven F. 59 Wuchter P. 44 Schillert A. 122 Starruß J. 89 van der Wekken A. J. 46 Schilling M. 59 Steiert B. 44, 79 van de Water B. 72, 76 Y Schilling M. 39, 44, 67, 68, Sten S.1 47 van Ham M. 111 Yakimovich A. 54 Stenzig J. 115 Vanlier J. 59, 72 Yepez V. 122 Schliemann-Bullinger M. 77 Stepath M. 44 Van Niekerk D. 90 Yilmaz Z. B. 39 Schmidt G. 92 Sterr M. 57 van Oudenaarden A. 28 Yin Y. 54 Schmidt L. 35 Stettler M. 41 Vatlitsov D. 75 Yugi K. 41, 78 Schmiedel J. M. 28 Steuer R. 68 Vaudry D. 91 Schnabel R. 122 Strålfors P. 44 v.d. Stel W. 72 Z Schneider A. 39 Strijkers G. 93 Veltman D. 70 Zahedi R. 99 Schneider C. 48 Strobel O. 101 Verschuuren M. 36 Zander J. 105 Schoeberl B. 45 Stys D. 53 Vieth B. 38 Zanghellini J. 83 Scholz M.1 107, 109, 110 Summers H. D. 47 Vignon-Clementel I. 43, Zanotelli V. 54, 65 Schönmeyer 95 Suzuki Y. 78 54, 88 Zehn D. 66 Schunkert H. 110 Swieringa F. 99 Vignon-Clementel I. E. 54 Zeilinger K. 48 Villiger T. K. 41 Zeissig S. 114 79, 80 Schuppert A. 96, 109, 101 Schuster S. 85, 122 T Vilne B. 110 Zeller T. 115 Schweinoch D. 78 Tanaka E. M. 35 Vinh J. 91 Zeller T. 122 Seddek A. 49 Tarlet G. 91 Vlaic S. 80, 122 Zerial M. 53 Sedlaczek O. 54 Taudt A. 46 Vlasov A. 80 Zerial M. 102, 114 Seehofer D. 48 Teif V. 98 Voit E. 81 Zerjatke T. 42, 55 Seehofer D. 39 Telfah A. 91 von den Driesch L. 122 Zhang F. 76 See V. 60 Tenzer S. 105 von Schönfels W. 114 Zhao D. 40 Seggewiss J. 119 Teplytska L. 104 Vvedenskaya O. 82 Zheng Y. 28 Segiova F. 114 Tessem M.- B. 52 Segovia-Miranda F. 53, 102 Teusink B. 82 W Zierer J. 61 Seifert M. 121 Teutonico D. 109 Wade J. 81 Zitzelsberger H. 120 Seifert S. 53 Thalheim T.1 42 Wagner M. 105 Ziv-Kenet A. 46 Sers, C. 28 Theis F. 43, 47, 55, 57, 58, 60, Wagner M.- C. 44, 67 Żulpo M.1 84 Sevimoglu T. 79 65, 70, 71, 80, 83, 101, 108, Waldherr S. 56 Shadrin A. 43 114, 118, 201 Warth A. 54 Shao C. 36 Thieffry D. 67 Wäsch M. 44, 79 Sharanek A. 107 Thiel C. 103 Watanabe S. 40 Shchekinova E. 89, 97 Thiele I. 31 Watanabe T. 40 She B. 67 Timens W. 46 Weaver R. J. 107 Shen Orr S. 46, 100 Timmer J. 6 9, 70, 72, 75, 77, Weber W. 69 Shoemaker J. 40 79, 87, 107 Weidemann A. 90 Sickmann A. 99 Timmer J. 39, 44, 67, 72 Wemmert C. 95 Sieber, A. 28 Tinneveld T. 82 Werner M. 113 Siebert H. 73 Tokarski C. 122 Westerhoff H. V. 111 Sieprath T. 36 Tönsing C. 87 Westermann F. 36 Sinkkonen L. 85 Trairatphisan P. 74 Weyßer F. 101 Sinnaeve P. 70 Tritschler S. 66 White M. 38, 60 Sison-Young R. 76 T. Sweeney E. 100 White M. 35, 57 Skogsberg J. 110 Turck C. 104 Wienhold S. 109 Smets M. 38 Wierling C. 43 Smith A. 42 U Wild P. S. 115 Snoep J. 90 Uda S. 78 Willemsen J. 39 Solari F. 99 Uhlenhaut H. 116 Wittig U. 90 Söllner J.1 121 Unger K. 120 Witzel, F. 28 Sonnenschein N. 81 Urlep Z. 40, 87 Witzenrath M. 109 Soumelis V.2 67 Ussar S. 63 Woetzel N. 105 Spector T. 61 Uvarovskii A. 98 Wolf A. 58 Spencer S. 47 126 Ziegenhain C. 38 Wolf F. A. 65 PP01 PP02 PP03 WC 3 1 4 1 Registration Desk 2 2Wardrobe 5 Bio Sys Net 3 3 3Catering 4BMFT Hörsaal A 5 Entrance PP07 PP04 Einsteinstraße PP06 PP05