Bioinformatics in the CDC Biotechnology Core Facility Branch Computational Lab Scott Sammons Kevin Tang Chandni Desai Sequencing Lab Mike Frace Missy Olsen-Rasmussen Marina Khristova Lori Rowe Genome Sequencing Lab sequencing platforms – current and upcoming AB 3730XL Roche 454 Titanium + Pacific Biosciences SMRT sequencer Illumina GA IIx Ion Torrent Personal Gene Machine Building 23 Server Room – Main ISLE 3 High Performance Computing Cluster (Aspen) • What is it? • 35 compute nodes each with 12 processor cores, 48GB of memory, and 2 Tesla 2050 GPU cards • Currently in the final stages of development in preparation for code-freeze and C&A • What can it do today? • 25 cluster applications are currently enabled for our phase-one deployment including MatLab, Geneious, Beast, Blast, and PacBio • Collaboration with NCI via IAA will GPU scientific applications even further • How fast is it? • By example, a Blast job that takes over 60 hours to complete on our old cluster takes 2 hours on the new cluster* 4 • *NOT GPU OPTIMIZED CODE Isilon • What is it? • High speed, scalable, and redundant Network Attached Storage • Currently in the process of being integrated with applications • Connected to both the CDC network and the Aspen HPC cluster utilizing Infiniband • What can it do today? • It provides user workspace for end-users and HPC applications • Solves the problem of being out of disk space on individual servers • What are we doing with it? • Data warehouse for all scientific equipment • Central network share for all scientific users • Integrating directly with ITSO’s Active Directory forest 5 Private Cloud • What is it? • Support science through front-end and back-end services • Implementation of virtualized infrastructure. • Currently in the process of being deployed. • What can it do today? • Provide test environments for scientific projects • Lay the foundation for hardware consolidation and migration • What are we doing with it? • Standardize platforms • Centralize management • Support ongoing growth within the scientific computing community while enabling science 6 Scientific Computing Infrastructure The Server Room • • • • • • • • • • • 2 Linux High Performance Computing Clusters (~40 nodes each) 1 Genomics Cluster 4 Solaris Servers 12 Stand-Alone Linux Servers 1 Stand-Alone Database Server 5 Stand-Alone Windows Servers Virtualized Cluster with 15 VMs 3 NAS Devices 2 Tape Libraries 2 Dedicated IP Subnets One C&A addressing all legacy production hardware (NCEZID) with several in-process for systems currently under development (NCIRD) 7 INFLUENZA GSL sequencing 2011 NCIRD Haemophilus influenzae Legionella pneumophila Legionella spp. Mycoplasma pneumonia Water cooling tower metagenomics Respiratory filter metagenomics Bat metagenomics NCEZID Vibrio cholera Vibrio spp Cyclospora Bacillus anthracis Listera Yersinia pestis Brucella spp. Klebsiella pneumonia Junin virus Rift Valley Fever virus Lujo virus Marburg virus CCHF virus Lassa Fever virus Clinical sample metagenomics Tick metagenomics Soil metagenomics CGH Guineaworm Taenia solium Angiostrongylus Sequencing: extended PCR Position of E-PCR overlapping amplicons A3 A1 End-L A2 D PO C A5 A4 A9 A7 A6 A8 A11 A10 E R K H ML I F N Q HindIII map A15 A13 A12 A14 A SJ A17 End-R A16 A18 B G Primers designed using VAR-BSH and VAC-CPN sequences Primers target genes involved in reproduction & host response Sequence sample: primers 40 sites, 1 enz. RFLP ~120 sites PCR uses minimal DNA amounts, often no need to grow virus PCR uses hifi expand long-template Taq & Pwo enzymes (Roche) First Pass Assembly: Seqmerge fold redundancy 16 12 8 4 Sequencing Assembly: Phred/Phrap/Consed Gene Prediction • Heuristic algorithm to assign quality scores to ORFs (from 1 to 100) • Quality scores are based on a number of factors including – Gene Predictions (glimmer, genemark, getorf) – Primary sequence homology to known genes (BLAST) – Presence of predicted promoter (MEME/MAST) – Size of predicted ORF – Presence of transcription terminal signals Visualizing Gene Predictions and Differences ORFs of CPVXs from 4 different clades ITR crm-D ITR 45 Smallpox Strains A. West African int. CFR ~10% C. Asian major CFR ~5 - 35% B. American alastrim minor CFR <1% C-1. non-WestAfrican-African int CFR ~10% C-2. non-WestAfrican African minor CFR <1% Unrooted tree phylogenetic relationships of ORF encoding the hemagglutinin protein Taterapox Camelpox Cowpox clade IV CPXV90_ger2 Variola AF375135 L22579 Ectromelia AY902256 Cowpox clade III (CPXV91_ger3) AY603355 AF377885 Cowpox clade II AF375086 VACLS1 Z99045 AY902297 Cowpox clade I Vaccinia AF375102 Monkeypox Next-Gen Diagnostic Sequencing Applications Shotgun / Paired-End Sequencing: random shearing of DNA, even sequence coverage over entire genome. ‘Massively parallel’ sequencing not only produces throughput, it provides sequences of potentially millions of individual molecules (instant cloning). By sequencing a PCR reaction it allows the detailed search for low expression quasi-species or mutations which may signal growing drug or vaccine resistance – a process called ultra-deep or amplicon sequencing. Example: clinical case of poxvirus infection with samples exhibiting a reduced sensitivity to an antiviral drug. Complex clinical, laboratory or environmental samples can be sequenced to provide a diagnostic ‘snapshot’ of the resident organisms - an approach called metagenomic sequencing. Examples: tissue culture, soil Shotgun / Paired-End Sequencing De novo Assembly • Newbler • CLCBio • Mira • Geneious • Velvet • Celera Reference Mapping • Newbler • CLCBio • Mosaik • Mira • Geneious • BWA Genome Assembly Visualization Genome Assembly Visualization Amplicon (deep) sequencing project Li, Damon - NCZEID/DVRD/PRB • Clinical case of progressive vaccinia infection from smallpox vaccination of an immune compromised patient • Pox antiviral ST-246 administered which targets pox gene F13L, a major envelope protein which mediates production of extracellular virus • Oral ST-246 given daily and vaccination site sampled over 3 week period A region of gene F13L was amplified from clinical samples, deep sequenced, and compared to the smallpox vaccine reference sequence (Acambis 2000) Control swab prior to ST-246 2 weeks after ST-246 T>A 943 C>T 869 3 weeks after ST-246 C>T 869 T>A 943 What is Metagenomics? • • Is the genomic study of DNA from uncultured microorganisms, generally from environmental samples Related • Metatranscriptomics • Metaproteomics Sample Coverage Rarefaction Curves Samples Wooley JC, Godzik A, Friedberg I, 2010 A Primer on Metagenomics. PLoS Comput Biol 6(2) Classification Techniques • Supervised Taxonomic Classification • Homology-based • Database searching by similarity (BLAST, SW) • BLAST, BLASTX: genbank, specialized DBs: NCBI-ENV-NT, NCBI-ENV-NR • Composition-based • N-mer frequency • Markov Models, Support Vector Machines (SVM), need training set • Unsupervised Taxonomic Classification • Clustering methods • SOM - self-organizing maps • PCA – principal component analysis Viral Metagenomic Pipeline (Wash U scripts implemented at CDC) Sample Collection Contigs, Reads DNA Library Construction Sequencing Remove redundant sequences Unique sequences Mask repetitive and low complexity seqs Good sequences BLASTN against Human Genome (e ≤ 1e-10) Basecalling Vector Trimming Assembly Non-human sequences BLASTX vs nr BLASTN vs nt BLASTN vs GB-viral Report Generation, Display in MEGAN, inspect top hits Software for Taxonomic Classification • • • • • • MEGAN – GUI interface for classification based on blast searches CARMA web-based classification using pFam database and HMMER alignment of protein families MG-RAST classification system utilizing protein encoding databases and several ribosomal DBs. Can analyze user provided datasets, web use only Geneious – commercial product NextGENe – commercial product Phymm, PhymmBL – composition based classification system Software for Comparative Metagenomics • • • Megan – can display two metagenome populations on the same phylogenetic tree, uses BLAST file as input STAMP – calculates statistical differences between sets of metagenomes XIPE-TOTEC – performs pairwise comparisons of every metagenome in the two sets, creates a distance matrix which is then used for clustering and PCA analysis to calculate statistical values of relatedness Megan Ugandan Outbreak Samples • 4 patients • Total RNA from patient sera • 2 samples per 454 run • ~ 565,000 reads/sample, avg length = 235nt • • • • Sequences were screened for random library amplication primers and low quality Assembled each run de novo using the 454 gsAssembler Performed a blastx database search using the assembled contigs (overnight) Visualized the blast output using MEGAN. MEGAN (MetaGenomeANalyzer) Ugandan Outbreak - results • • • Run1 - 5 contigs (out of 2463 > 100nt) matched YF virus, covering 98% of the genome (10,441 of 10,823bp) Mapped each sample from Run1 using an Ethiopian YF virus as reference. 3229 individual reads from Sample 1 indentified as YF. Run 2 – no YF reads found Phylogenetic analysis of yellow fever virus sequences Laura McMullan (DHPP/VSPB) Comparative Metagenomics – current work • • One 454 run Two samples • Sample 1 – ~578,000 reads, avg read length 438 bases • Sample 2 – ~550,000 reads, avg read length 425 bases • Total number of bases sequenced - ~488,000,000 Sample 1 – Rarefaction Curve Sample 1 Taxa tree (collapsed at the Order level) Comparison of Sample 1 and 2 Bioinformatics Tools • • • • • Bioinformatics Packages – EMBOSS – BioInquiry General Tools – Java/BioJava – Perl/BioPerl – BLAST Suite – BioEdit – GFFtoPS Genome Comparison/Alignment Tools – Mavid – Mauve – Clustal – Muscle Gene Prediction – Glimmer – GeneMark Assembly/Mapping Tools – 454 Suite – Mosaik Tools – Mummer – CLC Bio – BWA – Velvet – AHA (pacbio) • • • • Functional Annotation – Manatee Phylogenetics – Paup – Phylip – MrBayes – Beauti/Beast – MEGA – DnaSP Metagenomics – MEGAN – Galaxy – Carma In-House – WAMS – POCs/VOCs Challenges Data Management – image files are large (1 run ~25G) moving these files around the network is slow Assembly/Mapping Software – Some are provided with the instrument, but additional methods and algorithms are needed Finishing Tools – gap filling, primer design Visualization Tools – tools to graphically display contigs on reference sequence as well as genome multiple alignments Generic Robust Annotation Tools – Researchers need tools to intelligently choose predicted ORFs as genes, assign function, and submit to GenBank What are the weaknesses of current next-gen sequencers? Complicated and time consuming library preparation Requires micrograms of DNA to begin 3 days to prepare library Requires amplification of library Low copy number polymorphisms may be missed Emulsion PCR is an inefficient, time consuming, oily mess Potential to introduce PCR bias into sample Instruments require repetitive sequential ‘flows’ of reagents Repetitive flows of nucleotides, blocking/unblocking chemistry, washing out reaction byproducts all slow synthesis and hinder read-length Consumes liters of reagents ($) Repetitive flows and imaging extend sequence runs to days (or weeks) Pacific Bioscience SMRT sequencer (single-molecule sequencer) Ion Torrent Personal Gene Machine (solid-state sequencer) Nanopore sequencing Pacific Biosciences SMRT sequencer Sponsor: Influenza Research Agenda Pacific Biosciences SMRT Technology Individual ZMW with attached polymerase and DNA strand Laser excitation/detection volume glass ~ 50 nm SMRTcell = 160,000 ZMW Functional volume (red) is in zL! SMRTcell array = 1.5 million ZMW Nucleotide incorporation is a realtime data movie 100 ms Pacific Biosciences Advantages Read lengths of 1,000 – 10,000 bases 4 No reagent ‘flows’ =10-fold increase in sequencing speed Substitute reverse transcriptase for polymerase and sequence RNA directly Bacteria genomes sequenced in hours Sequence run costs 99$; take 15 minutes to complete 454 Sequencing • • • • DNA Library Prep emPCR Amplification Sequencing Data Analysis 454 Sequencing: DNA Prep • Nebulization – sheared with high pressure nitrogen to create fragments ~300-800 bases long • Repair Ends – double stranded pieces are purified, blunt ended, and phosphorylated • Adaptor Ligation – two different adaptors are ligated to the fragment, A and B – 44 bases long: 20 base PCR primer, 20 base sequencing primer, 4 base key – B fragment contain a biotin tag for immobilization – This forms 4 different strands A-A, A-B, B-A, B-B • Fragment Immobilization – These immobilized on streptavidin-coated magnetic beads, A-A strands will not bind and are washed away • Single-strand Isolation – bound fragments are denatured and the released strands (containing both an A and a B tag) form a single-stranded template DNA library 454 Sequencing: emulsionPCR Emulsion-based clonal PCR • Annealing – Fragments are annealed to primer tagged “catcher” beads – optimized to anneal a single strand to a single bead • Distribution in a water-oil-emulsion – the captured dna and beads along with amplication reagents are placed in a water-oil mixture – Each bead is captured in a “bubble” and creates its’ own small “micro-reactor” – thermocyled creating millions of copies of a single clonal fragment in individual “microreactors” – cleaned up and denatured 454 Sequencing: Sequencing by Synthesis • Bead Preparation - sequencing primer attached and polymerase and cofactors are added • Bead Deposition – beads are layered on a picotiter plate (wells are 44 μm), then enzyme beads and packing beads are added 454 Sequencing: Sequencing by Synthesis (cont.) • Sequencing – enzyme beads contain sulfurylase and luciferase, packing beads help keep reaction beads in position – a fluidics system delivers sequencing reagents, flowing the nucleotides one at a time in a specific order across the wells 454 Sequencing: Sequencing by Synthesis (cont.) • Sequencing – if a nucleotide is incorporated, a pyrophosphate is released which is converted to ATP by the sulfurylase – the ATP is hydrolyzed by the luciferase enzyme producing oxyluciferase and light – The light emission is measured with a CCD camera – light intensity indicates nucleotide incorporation 454 Sequencing: Sequencing by Synthesis (cont.) • Characteristics – Flow of the four nucleotides is repeated for one hundred cycles, resulting in average read length of 300-500 bases – system averages ~1,000,000 high quality wells – therefore, a typical run yields over 400 million high quality bases 454 Sequencing: Paired End Protocol