Next Generation Sequencing Bioinformatics Stephen Taylor Computational Biology Research Group

History Sanger Dominant for last ~30 years 1000bp longest read Based on primers so not good for repetitive or SNPs sites Next Generation Sequencing Much shorter reads, 25 to 300 bp Higher throughput Cheaper cost per Mb Single molecule sequencing (no cloning step) Since Jan 2008 more DNA sequenced than all previous years Sanger Dominant for last ~30 years 1000bp longest read Based on primers so not good for repetitive or SNPs sites Next Generation Sequencing Much shorter reads, 25 to 300 bp Higher throughput Cheaper cost per Mb Single molecule sequencing (no cloning step) Since Jan 2008 more DNA sequenced than all previous years Computational Biology Research Group

Hence We Need High Throughput Bioinformatics Computational Biology Research Group

Sanger Fred Sanger (1980) Dye-terminator sequencing PCR up DNA fragment Separate into 2 strands Polymerase elongates DNA Incorporation of fluorescence labelled ddNTP causes termination of elongation for each base Run DNA fragments on gel/capillary Peak generated for each base Fred Sanger (1980) Dye-terminator sequencing PCR up DNA fragment Separate into 2 strands Polymerase elongates DNA Incorporation of fluorescence labelled ddNTP causes termination of elongation for each base Run DNA fragments on gel/capillary Peak generated for each base Computational Biology Research Group

Illumina (Solexa) Computational Biology Research Group

Illumina (Solexa) Computational Biology Research Group

Illumina (Solexa) Computational Biology Research Group

Illumina (Solexa) Applications Resequencing Characterise different related species or strains Transcriptome analysis No chip/array required! random priming of RNA DNA methylation analysis sequencing bisulfite-converted DNA methylation-sensitive restriction digest enriched fragments Examine chromatin modifications Quantify in vivo protein-DNA interactions using the combination of chromatin immunoprecipitation and sequencing (ChIP-Seq) Resequencing Characterise different related species or strains Transcriptome analysis No chip/array required! random priming of RNA DNA methylation analysis sequencing bisulfite-converted DNA methylation-sensitive restriction digest enriched fragments Examine chromatin modifications Quantify in vivo protein-DNA interactions using the combination of chromatin immunoprecipitation and sequencing (ChIP-Seq) Computational Biology Research Group

Price Comparison Computational Biology Research Group

Processing and management Computational Biology Research Group

Assemble Data - Illumina Generates short reads (~35-75bp) Good for resequencing Difficult to do de novo assembly all but smallest organisms Generates short reads (~35-75bp) Good for resequencing Difficult to do de novo assembly all but smallest organisms Computational Biology Research Group

Mapping Illumina Reads Acquire and process images and convert to FASTQ* Get data Quality control** Map to genome Visualisation Post Processing Peak Finding SNP Calling * Not covered today Acquire and process images and convert to FASTQ* Get data Quality control** Map to genome Visualisation Post Processing Peak Finding SNP Calling * Not covered today Computational Biology Research Group

FASTQ TATACAATGCACTTAGTCATCCGCGTATCACTTTAT + IIIIIIIIIIIIIIIIIIGIIIIIIIIII4IIII:I 1.HWUSI-EAS100R the unique instrument name 2.6 flowcell lane 3.73 tile number within the flowcell lane 'x'-coordinate of the cluster within the tile 'y'-coordinate of the cluster within the tile 6.#0 index number for a multiplexed sample (0 for no indexing) /1 the member of a pair, /1 or /2 (paired-end or mate-pair reads TATACAATGCACTTAGTCATCCGCGTATCACTTTAT + IIIIIIIIIIIIIIIIIIGIIIIIIIIII4IIII:I 1.HWUSI-EAS100R the unique instrument name 2.6 flowcell lane 3.73 tile number within the flowcell lane 'x'-coordinate of the cluster within the tile 'y'-coordinate of the cluster within the tile 6.#0 index number for a multiplexed sample (0 for no indexing) /1 the member of a pair, /1 or /2 (paired-end or mate-pair reads only) Computational Biology Research Group

FASTQ format Quality Score ASCII representation of score for each base e.g. I Convert to ASCII e.g. 73 Minus Original Qphred= 40 See Quality Score ASCII representation of score for each base e.g. I Convert to ASCII e.g. 73 Minus Original Qphred= 40 See Computational Biology Research Group

Formats – warning! FASTQ format appears ‘standard’ but there are 3 types based on the probabilities of the base calls… Qphred = -10 x log10(error_prob) Qsolexa = -10 x log10(error_prob/(1-error_prob)) 1.Standard fastq: ASCII( Qphred + 33 ) 2.Illumina pre v1.3 : ASCII( Qsolexa + 64 ) 3.Illumina post v1.3: ASCII( Qphred+64 ) Option 3 should be the main one for the forseeable future! FASTQ format appears ‘standard’ but there are 3 types based on the probabilities of the base calls… Qphred = -10 x log10(error_prob) Qsolexa = -10 x log10(error_prob/(1-error_prob)) 1.Standard fastq: ASCII( Qphred + 33 ) 2.Illumina pre v1.3 : ASCII( Qsolexa + 64 ) 3.Illumina post v1.3: ASCII( Qphred+64 ) Option 3 should be the main one for the forseeable future! Computational Biology Research Group

Convert between formats Computational Biology Research Group Use sol2std2

Get Data May be supplied in a variety of formats.prb.txt files Contain probabilities for each base Some SNP callers use this Usually convert to FASTQ FASTQ Like FASTA but with quality score associated with each base May be supplied in a variety of formats.prb.txt files Contain probabilities for each base Some SNP callers use this Usually convert to FASTQ FASTQ Like FASTA but with quality score associated with each base Computational Biology Research Group

WTCHG If data is from WTCHG likely to get an E.g. wget the FASTQ file in the GERALD directory LD_ _johnb/s_2_sequence.txt.gz If data is from WTCHG likely to get an E.g. wget the FASTQ file in the GERALD directory LD_ _johnb/s_2_sequence.txt.gz Computational Biology Research Group

Processing reads - Illumina Mapping Tools MAQ Sanger Uses quality scores ELAND Comes with the machine and runs as standard Very fast NOVOALIGN Slower, more accurate Output option includes pairwise (handy for following up SNP calls) TOPHAT For RNA-Seq Can map slice junctions Mapping Tools MAQ Sanger Uses quality scores ELAND Comes with the machine and runs as standard Very fast NOVOALIGN Slower, more accurate Output option includes pairwise (handy for following up SNP calls) TOPHAT For RNA-Seq Can map slice junctions Computational Biology Research Group

Notes on Mapping What genome? Masking? Some tools disregard multiple maps e.g. ELAND Some tools map to one location and adjust probability score e.g. MAQ Can be confusing… For ChIP-Seq we normally use DNA heavily masked for repeats (simple/complex/ribosomal) What genome? Masking? Some tools disregard multiple maps e.g. ELAND Some tools map to one location and adjust probability score e.g. MAQ Can be confusing… For ChIP-Seq we normally use DNA heavily masked for repeats (simple/complex/ribosomal) Computational Biology Research Group

Databanks Indices We have many indexed databanks Under /databank/indices/ e.g. for maq ens_human_chrs/ ens_human_chrs_ucsc_rmfull_2/ ens_mouse_chrs/ ens_mouse_chrs_ucsc_rmfull/ ens_human_cdna/ ens_mouse_masked_chrs/ Indices for both maq and novoalign If an index you need is not there please ask – don’t make a local one in your account! We have many indexed databanks Under /databank/indices/ e.g. for maq ens_human_chrs/ ens_human_chrs_ucsc_rmfull_2/ ens_mouse_chrs/ ens_mouse_chrs_ucsc_rmfull/ ens_human_cdna/ ens_mouse_masked_chrs/ Indices for both maq and novoalign If an index you need is not there please ask – don’t make a local one in your account! Computational Biology Research Group

ChIP-Seq Pipeline Computational Biology Research Group ChIP-Sequencing Advantages Less DNA needed Not limited by micro-array content More precise site mapping Increased reads increases sensitivity Produces higher quality data

ChIP-Seq example NGS reads Map (maq) Peak pick (cisgenome) Extract sequences from features (Motif extract) MEME Weblogo

MAQ For simple runs use ‘easyrun’ option… nohup /proj/hts/bin/maq.pl easyrun -d maq.log In the main file is all.map To see the binary to something usable: maq pileup all.map > all.pileup These are quite large files… For simple runs use ‘easyrun’ option… nohup /proj/hts/bin/maq.pl easyrun -d maq.log In the main file is all.map To see the binary to something usable: maq pileup all.map > all.pileup These are quite large files… Computational Biology Research Group

Visualization all.map file converts to wig using CBRG custom tool maq wig all.map > all.wig Then we convert to GFF format using custom scripts all.map file converts to wig using CBRG custom tool maq wig all.map > all.wig Then we convert to GFF format using custom scripts Computational Biology Research Group

GFF format Gene Feature Format Developed at the Sanger Institute Format for describing features associated with DNA, RNA and Protein sequences Easy to parse More tools e.g. EMBOSS starting to use this as standard GFF3 is more standard and works best with GBrowse Gene Feature Format Developed at the Sanger Institute Format for describing features associated with DNA, RNA and Protein sequences Easy to parse More tools e.g. EMBOSS starting to use this as standard GFF3 is more standard and works best with GBrowse Computational Biology Research Group

##gff-version 3 chr3 src exon ID=exon00001 chr3 src exon ID=exon00002 chr3 src exon ID=exon00003 chr3 src exon ID=exon00004 chr3 src exon ID=exon00005 ##gff-version 3 chr3 src exon ID=exon00001 chr3 src exon ID=exon00002 chr3 src exon ID=exon00003 chr3 src exon ID=exon00004 chr3 src exon ID=exon00005 SOFA term Note ‘=‘

Wig binary files Scripts and modules to handle : UCSC wiggle format (1 column; 2 column; 4 column) or, gff3 binary (.wib) GMOD script wiggle_to_wigBinary.pl gff file Function: wiggle_to_wigBinary.pl variables (source / method / trackname / paths / input & output filenames ) command line to load binary / gff data into GBrowse (bp_seqfeature_load.pl + all variables: database name, filenames, paths etc) a conf file stanza - to display the loaded data construct an intermediate wiggle format file (....if input was gff3, maq binary)

Peak Calling Lots of algorithms to do this Problems with identifying a good cut off score Over and under prediction F-Seq Based on a training set of peaks identified by researcher in specific region Iterate over parameter space until achieve best TP/FP score cisgenome Uses IP and Non IP ChIP-Seq data, increases accuracy of predictions Lots of algorithms to do this Problems with identifying a good cut off score Over and under prediction F-Seq Based on a training set of peaks identified by researcher in specific region Iterate over parameter space until achieve best TP/FP score cisgenome Uses IP and Non IP ChIP-Seq data, increases accuracy of predictions Computational Biology Research Group

Motif Extraction Extract underlying DNA from peak calls Run using web based motif finders Weeder MEME May need to do successive rounds to find weaker motifs Extract underlying DNA from peak calls Run using web based motif finders Weeder MEME May need to do successive rounds to find weaker motifs Computational Biology Research Group

Quick note: SNP Calling Often finds errors in the PCR amplication step maq cns2snp (run during the easyrun option) SNPseeker Novoalign + CBRG script Worth trying all of the above! Often finds errors in the PCR amplication step maq cns2snp (run during the easyrun option) SNPseeker Novoalign + CBRG script Worth trying all of the above! Computational Biology Research Group

Molbiol Data Structure Analyse your data on deva.molbiol.ox.ac.uk CBRG set up /proj/hts/data/ Suggested structure: batch/ fastq/ dbname/ Contact us if you want a GBrowse database for your data Analyse your data on deva.molbiol.ox.ac.uk CBRG set up /proj/hts/data/ Suggested structure: batch/ fastq/ dbname/ Contact us if you want a GBrowse database for your data Computational Biology Research Group

Future Problem In depth analysis after mapping = bottleneck Need to empower the users to do their own analysis Solution Makefiles for bulk data analysis Allow access to NGS data via GBrowse ‘workbench’ GBrowse plugins to export data to other tools Galaxy looks promisinghttp://main.g2.bx.psu.edu/ Problem In depth analysis after mapping = bottleneck Need to empower the users to do their own analysis Solution Makefiles for bulk data analysis Allow access to NGS data via GBrowse ‘workbench’ GBrowse plugins to export data to other tools Galaxy looks promisinghttp://main.g2.bx.psu.edu/ Computational Biology Research Group