Massive Parallel Sequencing Kun Huang, PhD Department of Biomedical Informatics OSU CCC Bioinformatics Shared Resources

Introduction High throughput sequencing – a new paradigm Applications Solexa SOLiD 454/Roche Genome Sequencer Applications Genome sequencing microRNA screening Gene expression ChIP-seq Genome-worth of sequence, terabytes of data

What is ChIP-Sequencing? ChIP-Sequencing is a new frontier technology to analyze protein interactions with DNA. ChIP-Seq Combination of chromatin immunoprecipitation (ChIP) with ultra high-throughput massively parallel sequencing Allow mapping of protein–DNA interactions in-vivo on a genome scale

Workflow of ChIP-Seq Mardis, E.R. Nat. Methods 4, 613-614 (2007)

Workflow of ChIP-Seq

ChIP-seq Challenges: Millions of segments Mapping to genome Visualization Peak detection Data normalization …

Johnson et al, 2007 ChIP-Seq technology is used to understand in vivo binding of the neuron-restrictive silencer factor (NRSF) Results are compared to known binding sites ChIP-Seq signals are strongly agree with the existing knowledge Sharp resolution of binding position New noncanonical NRSF binding motifs are identified

Robertson et al, 2007 ChIP-Seq technology used to study genome-wide profiles of STAT1 DNA association STAT1 targets in interferon-γ-stimulated and unstimulated human HeLA S3 cells are compared The performance of ChIP-Seq is compared to the alternative protein-DNA interaction methods of ChIP-PCR and ChIP-chip. 41,582 and 11,004 putative STAT-1 binding regions are identified in stimulated and unstimulated cells respectively.

Why ChIP-Sequencing? Current microarray and ChIP-ChIP designs require knowing sequence of interest as a promoter, enhancer, or RNA-coding domain. Lower cost Less work in ChIP-Seq Higher accuracy Alterations in transcription-factor binding in response to environmental stimuli can be evaluated for the entire genome in a single experiment.

Bioinformatics

Sequencers Solexa (Illumina) 454 Life Sciences (Roche Diagnostics) 1 GB of sequences in a single run 35 bases in length 454 Life Sciences (Roche Diagnostics) 25-50 MB of sequences in a single run Up to 500 bases in length SOLiD (Applied Biosystems) 6 GB of sequences in a single run

Illumina Genome Analysis System 8 lanes 100 tiles per lane

Sequencing

Sequencer Output Quality Scores Sequence Files

Sequence Files ~10 million sequences per lane ~500 MB files

Quality Score Files Quality scores describe the confidence of bases in each read Solexa pipeline assigns a quality score to the four possible nucleotides for each sequenced base 9 million sequences (500MB file)  ~6.5GB quality score file

Bioinformatics Challenges Rapid mapping of these short sequence reads to the reference genome Visualize mapping results Thousand of enriched regions Peak analysis Peak detection Finding exact binding sites Compare results of different experiments Normalization Statistical tests

Mapping of Short Oligonucleotides to the Reference Genome Mapping Methods Need to allow mismatches and gaps SNP locations Sequencing errors Reading errors Indexing and hashing genome oligonucleotide reads Use of quality scores Use of SNP knowledge Performance Partitioning the genome or sequence reads

Mapping Methods: Indexing the Genome Fast sequence similarity search algorithms (like BLAST) Not specifically designed for mapping millions of query sequences Take very long time e.g. 2 days to map half million sequences to 70MB reference genome (using BLAST) Indexing the genome is memory expensive

SOAP (Li et al, 2008) Both reads and reference genome are converted to numeric data type using 2-bits-per-base coding Load reference genome into memory For human genome, 14GB RAM required for storing reference sequences and index tables 300(gapped) to 1200(ungapped) times faster than BLAST

SOAP (Li et al, 2008) 2 mismatches or 1-3bp continuous gap Errors accumulate during the sequencing process Much higher number of sequencing errors at the 3’-end (sometimes make the reads unalignable to the reference genome) Iteratively trim several basepairs at the 3’-end and redo the alignment Improve sensitivity

Mapping Methods: Indexing the Oligonucleotide Reads ELAND (Cox, unpublished) “Efficient Large-Scale Alignment of Nucleotide Databases” (Solexa Ltd.) SeqMap (Jiang, 2008) “Mapping massive amount of oligonucleotides to the genome” RMAP (Smith, 2008) “Using quality scores and longer reads improves accuracy of Solexa read mapping” MAQ (Li, 2008) “Mapping short DNA sequencing reads and calling variants using mapping quality scores”

Mapping Algorithm (2 mismatches) GATGCATTGCTATGCCTCCCAGTCCGCAACTTCACG GATGCATTG CTATGCCTC CCAGTCCGC AACTTCACG seeds GATGCATTG CTATGCCTC CCAGTCCGC AACTTCACG ......... Exact match Genome Indexed table of exactly matching seeds Approximate search around the exactly matching seeds

Mapping Algorithm (2 mismatches) Partition reads into 4 seeds {A,B,C,D} At least 2 seed must map with no mismatches Scan genome to identify locations where the seeds match exactly 6 possible combinations of the seeds to search {AB, CD, AC, BD, AD, BC} 6 scans to find all candidates Do approximate matching around the exactly-matching seeds. Determine all targets for the reads Ins/del can be incorporated The reads are indexed and hashed before scanning genome Bit operations are used to accelerate mapping Each nt encoded into 2-bits

ELAND (Cox, unpublished) Commercial sequence mapping program comes with Solexa machine Allow at most 2 mismatches Map sequences up to 32 nt in length All sequences have to be same length

RMAP (Smith et al, 2008) Improve mapping accuracy Possible sequencing errors at 3’-ends of longer reads Base-call quality scores Use of base-call quality scores Quality cutoff High quality positions are checked for mismatces Low quality positions always induce a match Quality control step eliminates reads with too many low quality positions Allow any number of mismatches

Mapped to a unique location Map to reference genome Mapped to a unique location Mapped to multiple locations No mapping Low quality 7.2 M 1.8 M 2.5 M 0.5 M 12 M 3 M Quality filter

Bioinformatics Challenges Rapid mapping of these short sequence reads to the reference genome Visualize mapping results Thousand of enriched regions Peak analysis Peak detection Finding exact binding sites Compare results of different experiments Normalization Statistical tests

Visualization BED files are build to summarize mapping results BED files can be easily visualized in Genome Browser http://genome.ucsc.edu

Visualization: Genome Browser Robertson, G. et al. Nat. Methods 4, 651-657 (2007)

Visualization: Custom 300 kb region from mouse ES cells Mikkelsen,T.S. et al. Nature 448, 553-562 (2007)

Visualization Huang, 2008 (unpublished)

Huang, 2008 (unpublished)

Bioinformatics Challenges Rapid mapping of these short sequence reads to the reference genome Visualize mapping results Thousand of enriched regions Peak analysis Peak detection Finding exact binding sites Compare results of different experiments Normalization Statistical tests

Peak Analysis Peak Detection ChIP-Peak Analysis Module (Swiss Institute of Bioinformatics) ChIPSeq Peak Finder (Wold Lab, Caltech)

Peak Analysis Finding Exact Binding Site Determining the exact binding sites from short reads generated from ChIP-Seq experiments SISSRs (Site Identification from Short Sequence Reads) (Jothi 2008) MACS (Model-based Analysis of ChIP-Seq) (Zhang et al, 2008)

Bioinformatics Challenges Rapid mapping of these short sequence reads to the reference genome Visualize mapping results Thousand of enriched regions Peak analysis Peak detection Finding exact binding sites Compare results of different experiments Normalization Statistical tests

Compare Samples Huang, 2008 (unpublished)

Compare Samples Fold change HPeak: An HMM-based algorithm for defining read-enriched regions from massive parallel sequencing data Xu et al, 2008 Advanced statistics

QUESTIONS?