Systems Drug Design Hiroaki Kitano The Systems Biology Institute Sony Computer Science Laboratories, Inc. Okinawa Institute of Science and Technology Department of Cancer Systems Biology, The Cancer Institute We cure & We care Systems Drug Design Coral Reef Systems Biology SBI Strategy • Innovation in drug design and systems medicine • Faster social and business impacts • Global strategy (Singapore, India, Shanghai) • Rolling out business operations SBI Collaborative Drug Pipelines An early stage list Discovery Preclinical Phase-I Phase-II Phase-III X-7CD TB with CSIR India Breast cancer Influenza CNS (SZ, PD, AD) JSPS & OIST-SBI Project With ERATO Kawaoka Project PD-I program is with Univ. Luxembourg Cardiovascular system related Discovery Phase: Identification of possible molecular targets for a given disease Translational Phase: (1) Given a candidate compound, identify what is the best disease subtype (2) Given a candidate compound and target disease, find what other drugs to be used in combination Software Platform Computational platform for systems drug discovery Target Market Segments Personalize Premiere Medical and Wellness Services High income bracket Comprehensive medical and wellness service Healthcare version of Private Bank Affordable medical services Mass market Quality service at low cost Treatments for each patient cluster Humanitarian Medical Support Medicare for Bottom Billions Cost and Access Cost Gefitinib (Iressa: AZ) Indication: Non-small cell lung cancer Efficacy: For patients with EGFR mutation, overall response rate was 75% EGFR mutation in 25% of Japanese patients 2% in U.S.A. Side effects: Interstitial pneumonia (IP) 5.8% of Japanese patients with 50% mortality rate Distribution of mutations in NSCLC Sharma, et al., Nature Reviews Drug Discovery, April, 2010 9 10 11 12 11 Target Market Segments Personalize Premiere Medical and Wellness Services High income bracket Comprehensive medical and wellness service Healthcare version of Private Bank Affordable medical services Mass market Quality service at low cost Treatments for each patient cluster Humanitarian Medical Support Medicare for Bottom Billions Cost and Access Cost TB: A Disease Neglected Robustness An ability of the system to maintain its functions even under external and internal perturbations Cancer Robustness • Major sources of robustness – Feedback loops and crosstalks within cell – Heterogeneity of mutations • Due to point mutations, mitotic recombination, anueploidy • – Host-Tumor Entrainment • Hypoxia Inducible Factors, microenvironment remodeling • Self-extending symbiosis: Cell fusion, chromosome intake, macrophage, etc. Kitano, Nature Rev. Cancer, 4, 227-35 2004 Kitano, Nature, 426, 125 2003 Kitano and Oda, Biological Theory, 2006 Intra-tumour heterogeneity (Colorectal Cancer) Baisse, et al., Int. J. Cancer, 93, 346-352, 2001 Robustness Trade-offs Systems that are optimized for certain perturbations inevitably entail extreme fragility elsewhere. Kitano, Nature Reviews Genetics, 2004, Kitano, Molecular Systems Biology, 2008 Cset and Doyle, Science, 2002 Robustness-Fragility trade-offs in control theory (negative feedback) Bode Theorem (Bode 1945) Cset & Doyle, Science, 2002 Yi, et al., Basic control theory for biologists, 2002 Kitano, Mol. Syst. Biol., 2007 Collateral Sensitivity Resistance Fragility Multiple genes are involved in many diseases Goh, et al., PNAS 2007 25%~30% of Hubs are Involved in cancer Inhibiting HUBs may cause serious side-effects Goh, et al., PNAS 2007 N > 36 35 > N > 6 5>N Budding Yeast PIN Human PIN Hase et al., PLoS Computational Biology, Oct. 30. 2009 Hase et al., PLoS Computational Biology, 30 Oct 2009 Internet Router Topology Human PPI 27 Targets of FDA Approved Drugs Hase et al., PLoS Computational Biology, 30 Oct 2009 Significance (connection, frequency, etc) Long Tail Distribution (log-linear graph) Head Tail Rank EMBO Symposium Combinatorial High Throughput Screening Multicomponent therapeutics that prevent proliferation of fluconazole-resistant C. albicans Borisy, Alexis A. et al. (2003) Proc. Natl. Acad. Sci. USA 100, 7977-7982 Copyright ©2003 by the National Academy of Sciences Chlorpromazine, an antipsychotic agent, and pentamidine, an antiprotozoal agent, together selectively prevent tumor cell growth in vitro and in vivo Efficacy Phase 1/2A Stage Borisy, Alexis A. et al. (2003) Proc. Natl. Acad. Sci. USA 100, 7977-7982 Copyright ©2003 by the National Academy of Sciences Drug Price Possible reduction of drug price – Taxol : 100mg 43768 • Bristol-Myers Squibb, Paclitaxel – Contomin : 100mg 9.2 • Tanabe-Mitsubishi, Chlorpromazine – Benanbax : 100mg • Sanofi-Aventis, Pentamizine 2824 Rhabdoid tumour xenograft Rhabdomyosarcoma xenograft Rapamycin: 5mg/kg daily for 5 consecutive days / week = MTD Cyclophosphamide: 150mg/kg daily = MTD Combination = MTD for both Kummar, et al., Nature Reviews Drug Discovery, Nov. 2010 Originally from Houghton, et al., Mol. Cancer Ther. 9, 101-112 (2010) Differential Robustness Screening Robustness-based target candidate selection Differential Robustness Model A (wt) Model B (mutant) k4 cyclin degradation k6 k1 cyclin synthesis cyclin degradation k1 cyclin synthesis Morohashi, et al., J. Theor. Biol., 216, 19-30 2002 Moriya, Shimizu-Yoshida, Kitano, PLoS Genetics, 14 July 2006 Upper-bound dosage of cell cycle related genes -leucine -uracil Moriya, Shimizu-Yoshida, Kitano, PLoS Genetics, 14 July 2006 gTOW-6000 Genome-wide gTOW Collection Computational approach for combinatorial problems An example workflow of model-driven biology Ghosh et al., Nature Reviews Genetics, Nov. 2011 42 Deep Curation EGF Receptor Cascade Oda, et al. Molecular Systems Biology, 2005 47 CellDesigner Modeling tool for biochemical and gene-regulatory network + + + http://celldesigner.or “PAYAO” Community Tagging System to SBML models • A community tool to work on the same pathway models simultaneously, insert tags to the specific parts of the model, exchange comments, record the discussions and eventually update the models accurately and concurrently. • Reads SBML models, display them with CellDesigner PAYAO: SBML Models Tagging System Large Scale Network Map for Breast Cancer Tamoxifen Resistance Molecular interactions of ERα interactions in MCF-7 cell lines curated from literature and represented in SBML format (CellDesigner 4.0.1) Signaling network interactions 142 proteins 256 reactions 126 complexes ~200 publications Transcriptional activity of ERα Reconstructed phosphoproteomics network Expression profile based focusing of genes and pathways Oyama, et al., JBC 286 (1) 818-829, 2011 52 Dynamic model construction • • Dynamic model encompassing major players of the ligand-independent ER activation Model adapted from existing ERBB network models (Chen et.al 2009, Wolf et.al 2007, Birtwistle et.al 2007) Model abstracts ERBB dimerization states (Birtwistle et.al 2007) Growth-factor PI3K mediated pathway HRG 90 state variables 80 reactions (ODE) Akt ~150 parameters ERBB Dimers Adapter molecules PI3K-AKT –Erα crosstalk Ras MAPK Erα crosstalk ERα@167 Raf Mek Erk ERα@118 79 Experimental results Molecular components of MAPK and PI3K-Akt pathways are highly phosphorylated compared to WT cells ERα @167 characterized by 10-fold amplification in phosphorylation in TamR cells 10-fold amplified phosphorylation Simulation reproducing experimental results Models based dynamics of the molecular components identify elevated phosphorylation states, particularly for ERα @167 Sensitivity Analysis: ERα @167 phosphorylation sensitive to PI3K-Akt arm 10-fold amplified phosphorylation 80 PI3K ERα Akt 1 2 Phosphorylation of Akt Activation of ERα 3 De-phosphorylation of ERα How to develop high precision simulation? 57 Comparison of robustness profile and a computational model Possible causes of differences • Treatment of Paralogues (CLB1-2, CLB3-4, CLB5-6 etc.) • Treatment of Stoichiometric Inhibitor (Clbs-Sic1, Esp1Pds1, Net1-Cdc14) Moriya, Shimizu-Yoshida, Kitano, PLoS Genetics, 14 July 2006 Cdc14, Net1 Esp1, Pds1 are all essential genes Kaizu, Moriya, Kitano, PLoS Genetics, 2010 Cleavage of Mcd1 by Caspase-like Protease Esp1 Promotes Apoptosis in Budding Yeast Hui Yang, Qun Ren, and Zhaojie Zhang, Mol. Biol. Cell, Vol. 19, Issue 5, 2127-2134, May 2008 Kaizu, Moriya, Kitano, PLoS Genetics 2010 Kaizu, Moriya, Kitano, PLoS Genetics 2010 Budding Yeast Cell Cycle and Signaling 64 Kaizu, Moriya, Kitano, PLoS Genetics 2010 Kaizu_Figure S3 A B Chen’s model C D ESP1-op ESP1-op, PDS1-op Amount (unit) Esp1total Amount (unit) Wild type Time (min.) Time (min.) Cell mass Esp1active Time (min.) Esp1total Kaizu, Moriya, Kitano, PLoS Genetics 2010 A Transport model Esp1active B C D ESP1-op Time (min.) Time (min.) ESP1-op, PDS1-op Amount (unit) Esp1total Amount (unit) Wild type Cell mass Esp1active Time (min.) Esp1total Kaizu, Moriya, Kitano, PLoS Genetics 2010 A Esp1 phosphorylation model C ESP1-op Amount (unit) Wild type Amount (unit) B Time (min.) Cell mass Time (min.) Esp1active Esp1total Kaizu, Moriya, Kitano, PLoS Genetics 2010 Kaizu_Figure S6 Pds1 phosphorylation model Phosphorylation is prevented ESP1-op Amount (unit) Amount (unit) Wild type Time (min.) Time (min.) Cell mass Esp1active Esp1total Kaizu, Moriya, Kitano, PLoS Genetics 2010 WT Pds1 or Esp1 deletion Pds1 and Eps1 deletions are both lethal, thus effects Clb2 based buffering cannot be observed Clb2 deletion No phenotype if Esp1:Psd1 balance is kept normal Clb2 deletion + Esp1 over-expression Comparison of robustness profile and a computational model Possible causes of differences • Treatment of Paralogues (CLB1-2, CLB3-4, CLB5-6 etc.) • Treatment of Stoichiometric Inhibitor (Clbs-Sic1, Esp1Pds1, Net1-Cdc14) Moriya, Shimizu-Yoshida, Kitano, PLoS Genetics, 14 July 2006 Software problems • Software for biomedical research is the critical components for success of research • Nobody can develop entire software systems alone • However ….. – Tools are developed independently – Different GUI, different operating procedure, different APIs, etc. – Need to launch tools independently – No direct data sharing, etc • Inter-operability is missing!!!! • Extra work needed for users and developers (C) Hiroaki Kitano, 2010 *** LIMITED CIRCULATION *** Data and Knowledge base Problems • Too many fragmented DBs and KBs. • Inconsistency/maintenance/errorcorrection • Users are forced to integrate by them self. • Poor feedback mechanism exists that prevents DB/KB to improve their quality (C) Hiroaki Kitano, 2010 *** LIMITED CIRCULATION *** The Garuda Alliance • Developer Benefits – Consistent GUI, APIs, and other development framework – Enables efficient and quality software development – Effective dissemination of tools and resources • User Benefits – – – – One Stop Service A consistent user experience Highly interoperable software tools Stable software platform Garuda Vision A common platform of tools that supports applications Garuda modules can be tailored to leverage functions across disparate tools which otherwise do not inter-operate, while integrating public domain knowledge spread across isolated databases Payao Merge iPath ARENA3D KLEIO CellDesigner Modeling tool for biochemical and gene-regulatory network + + + http://celldesigner.or Inheriting in silico IDE Tight integration with CellDesinger Supports ISML, SBML, etc. Garuda compliant 28 Garuda Vision A common platform of tools that supports applications Garuda modules can be tailored to leverage functions across disparate tools which otherwise do not inter-operate, while integrating public domain knowledge spread across isolated databases Payao Merge iPath ARENA3D KLEIO 40 www.garuda-alliance.org HD-Physiology Project Heart model ADME/PK model drug Doze, patterns, etc. Inter-layer interactions ADME/PK Action potential Genetic Polymorphism Intra-cellular interaction Cellular model Electrophysiology Molecular level models ADME/PK Loosely coupled real-time computing Inter-cellular dynamics Action potential Electrophysiology Off-line computing and visualization MD/BD Off-line computing and parameter integration 115 Possible application of cell based toxicology Prediction of QT elongation QT elongation is one of the major cause of drug withdrawal. HERG channel is the main target of QT elongation. Prediction of Hepatotoxicity Liver takes central role in the clearance and transformation of chemicals Step 1 oxidation, reduction, hydrolysis, hydration Step 2 transferase Hepatotoxicity means the liver damage induced by chemicals. Hepatotoxicity is one of the major cause of drug withdrawal. EMBO Symposium Kitano, et al., Nature chemical biology, 2011 EMBO Symposium 88 The First Molecular Interaction Map of TB OSDD-SBI collaboration Kitano, Ghosh, Matsuoka, Nature Chemical Biology, May 2011 Kitano, Ghosh, Matsuoka, Nature Chemical Biology, May 2011 Theories: Robustness, etc Computational Modeling & Simulations Data Analysis Technology Platform Goal-driven project management and decision making Universal approach for personalized medicine and unmet medical needs 94