Bioinforma)cs Series: Designing An NGS Study For My ... · Comprehensive Molecular Characterization...

Bioinforma)csSeries:DesigningAnNGSStudyForMyBiological

Ques)on

YussanneMaGenomeSciencesCentre

GenomeSciencesCentre

Sequencing5IlluminaHiseqX4IlluminaHiSeq25002NextSeq5003IlluminaMiseq1Life3730xl1500librariespermonth>80Tbasespermonth

Compute2secureddatacentresComputeclusters:800nodes,24,000

hyper-threadedcores48-384GBRAMpernodeHighmemory(1.5TBRAM)computers>11Petabyteson-linediskstorage

Engagedinover50ongoingprojectsandcollaboraYonsfromexperimentaldesign

todatainterpretaYon

Whyexperimentaldesignisimportant

Analysis and interpretation of sequencing data is completely dependent on everything upstream

Sample quality, sequencing and type of sequencing matter

The area being sampled matters

Bioinformatic corrections can be made but it’s always best to plan ahead

Outline

GenomesequencingTranscriptomesequencingIntegraYveapproachesOthertechnologies

Outline

GenomesequencingGenotypingarraysExomeandcustomcaptureAmpliconWholegenomesequencingPopulaYonsizeandcontrolsFactorsaffecYngqualityofvariantcalls

Genomesequencingoverview

SNP array

Custom Capture

Exome Capture

WGS

Genome Intron Exon

There are many ways to subsample the genome The cost trade-off is between area covered and depth Sometimes the genome can be overkill

Genotypingarrays

SamplingofthegenomeatlocaYonsofknownsinglenucleoYdepolymorphismsusingintensityprobes

Usedfor:Studyingcommonvariantsinlargenumberofcasesandcontrols

LimitaYons:Cannotbeusedforcallingofrareornovelvariants,orstructuralvariants.ResoluYonsforcopynumbervariaYonarelow

Exampleproject:Genome-wideassociaYonstudytolookforinheritedcancersuscepYbilityloci

Intron Exon

Exomeandcustomcapture

Probesareusedtocaptureallexons,oraspecificsetofgenomicregions

Usedfor:Studyingonlycodingchanges,orathosefoundinthepre-definedareaofinterest

LimitaYons:Cannotcallvariantsoutsideofcapturearea.Copynumberandstructuralvariantsaredifficulttocall

Exampleproject:Discoveryofrecurrentlymutatedgenesinlargecohort,clinicalpanel

Intron Exon

AmpliconandSangersequencing

Primersaredesignedthatspanagenomicevent,orsequenceacrossanevent.

Usedfor:Determiningthepresenceorabsenceofspecificevents(SNVs,indels,SVs)

LimitaYons:Cannotbeusedfordiscovery,needexactbreakpointsinmostcases

Exampleproject:OrthogonalverificaYonofputaYveeventdiscoveredbyWGStobenchmarktools,determiningpresenceofmetastaYcfusioneventinprimarysample.

Intron Exon

X

Wholegenomesequencing

Usedfor:FullcharacterizaYonofgenomeincludingNovelgenesandeventsSNVsandindelsnotseeninthepopulaYonPrivateeventsinrecurrentgenesorpathways

ComplexeventsCopynumberStructuralvariants

GenomiclandscapeofapopulaYonMutaYonsignatures

LimitaYons:SamplesizeanddepthduetocostBestusedforstudieswithnoaprioriknowledgeof

populaYonsamples,indepthstudyofsinglepaYenttumour

Genomesequencing:summary

Technology SNVs CNVs SVs Muta)onalburden

Muta)onallandscape

WGS +++ +++ +++ +++ +++

Exome +++(codingonly)

+(coding)

+(coding)

++

-

Customcapture +++(ontarget)

+(ontarget)

+(ontarget)

+

-

Genotypingarray Specificeventsonly + - - -

Amplicon Specificeventsonly - Specific

eventsonly - -

SangerSpecific

eventsonly

- Specificeventsonly - -

PopulaYonsizeandcontrols3milliongermlinevariants,10,000-100,000somaYcvariantsonaveragepersample

Raw variant calls

High quality variants

Not in controls

Functionally important

PopulaYonsizeandcontrols

GWAS:LargesamplesizeneededtoachievestaYsYcalsignificance,1:1casesandcontrols

Raredisease:Sequencingofparentsrevealspafernsofinheritance,sequencingofunaffectedrelaYveshelpstofilteroutpassengers

SomaYcvariants:MatchednormalisneededtofilteroutpassengermutaYons

Factorsaffec)ngqualityofvariantcalls:sequencingdepth

Why30Xgenome?Ruleofthumb:ittakes3-10highqualityreadstocallavariantNeedtoaccountforvariablecoverage,evennessofcoverage,tumour

content,ploidy

Typeofvariant Depthneeded

Germline,diploid 30X

Tumour>70%tumourcontent 30-40X

Tumour40-70% 40-60X

Lowtumourcontent,subclonalevents >100X

Factorsaffec)ngqualityofvariantcalls:sequencingdepth

FFPEvsFreshfrozenAllofourprotocols(WGS,RNA,miRNA)canberunonFFPE

samples,buttheymayresultinslightlyloweryieldanddiversity,andahigherfalseposiYverateforSNVandSVdetecYon,aswellasnoisierCNVcalls

Fresh FFPE

Factorsaffec)ngqualityofvariantcalls:sampletype

Outline

TranscriptomesequencingRNAsequencingmiRNAsequencingBatcheffects

RNAsequencing

RibosomaldepleYonvs.polyAselecYon

no ribosomal RNA captured non-polyadenylated transcripts are captured lower minimum input requirement higher intergenic and intronic content

higher ribosomal RNA content only polyadenylated transcripts are captured higher minimum input requirement lower intergenic and intronic content

RNAsequencing

Usedfor:Gene,exonandisoform-levelquanYficaYonQuanYfyingexpressionofgenomicevents(SNVs,SVs)DetecYngnoveltranscriptsDetecYngRNAeditsDifferenYalexpressionbetweengroups(condiYon/Yssue/

tumourtype)toidenYfyexpressionmarkersCorrelaYonandclusteringofsamplesbygeneexpressionto

idenYfysubgroups

RNAsequencingDifferenYalexpressionbetweentumourgroupsResultsaremoredifficulttointerpretwithlowsamplesize

RNAsequencingHierarchicalclusteringofmedulloblastomasamplesbygeneexpressionSamplesclusterbysubgroupWithinasubgroupsamplesclusterbypaYenttypeMostinformaYvewithlargesamplesizeanddetailedcovariateseg.clinicaldata

Sequencingdepth

Gene diversity for UHR control at different levels of downsampling Diversity does not tend to saturate

SequencingdepthNumberofreadsperlibrary 200M 120M 60M 40M

Numberofgenesat1X 23,000 20,000 18,000 15,000

Numberofgenesat10X 14,000 12,000 10,000 5,000

ExpressionquanYficaYon ++ ++ ++ ++

DifferenYalexpression ++ ++ ++ ++

KnowntranscriptquanYficaYon ++ ++ ++ ++

DetecYonofstructuralvariantswithgenepartnersorbreakpointsspecified

++ ++ ++ +

DetecYonofSNVsandsmallindelswithknowncoordinates

++ ++ ++ ++

DenovoSNVcalling ++ ++ + -

Denovostructuralvariantcalling ++ + Alignmentbasedonly

-

miRNAsequencingUsedfor:QuanYficaYonofmiRNAexpressionDifferenYalexpressionandexpressionclusteringCorrelaYonwithgeneexpressiontoidenYfytargets

Using miRNA to determine a signature for prognosis miRNA clustering is found to be more sensitive to subgroups Search space is smaller and cost of sequencing is lower May be easier to translate into clinical test

Sequencingdepth

miRNA diversity vs number of reads aligned to miRNA Two failed samples on far left Saturation between 2 and 4 million reads

Batcheffects

Samples cluster by protocol, so clustering is difficult to do across multiple protocols Sample sets sequenced using different protocols are best used as validation, or for meta analysis Batch effect correction is the most effective with technical replicates

Outline

IntegraYveapproachesGenomeandtranscriptomesequencingClonalevoluYonexperimentIntegraYveanalysistostudy‘darkmafer’incancer

IntegraYveapproaches:genomeandtranscriptomeRNAseq provides orthogonal validation of genomic events Combined approach improves specificity, and can identify/confirm alternative splicing and elucidate the effects of genomic events on transcription

Alternative splicing at M1 is identified in the structural variant analysis of RNA and DNA, and gene expression data confirms exon 9 skipping event

IntegraYveapproaches:ClonalevoluYon

Tumour

Mousemodel

Mousemodel

Mousemodel

WGS to identify SNVs common to tumour and xenografts

Mousemodel

Mousemodel

Mousemodel

Mousemodel

Mousemodel

Mousemodel

Mousemodel

Mousemodel

Mousemodel

Custom panel designed from WGS calls and literature

Clonal evolution over multiple timepoints and treatment events

Timepoint 1 Timepoint 2 Timepoint 3

Rx1 Rx2

IntegraYveapproaches:‘darkmafer’

Comprehensive Molecular Characterization of Papillary Renal-Cell Carcinoma, The Cancer Genome Atlas Research Network, N Engl J Med 2016; 374:135-145, January 14, 2016

miRNA, RNA and WGS and methylation sequencing identify multiple mechanisms in which CDKN2A function is disrupted in papillary renal-cell carcinoma

Outline

OthertechnologiesMicrobialanalysisdenovogenomeassemblySingle-cellsequencingEpigenomicsImmunogenomics

Microbialanalysis

16Ssequencing:IdenYficaYonandquanYficaYonofknownbacterialspecies.Usefulforsurveyoflargenumberofsamples

Shortreadsequencing(shallow):RapidclassificaYonofknownmicrobialspeciesinmetagenomicsamples

Wholegenomeandtranscriptomesequencing:MicrobialexpressionandgenomeintegraYonintumoursamples

Denovogenomeassembly

Shortreadsequencingat~30XissufficientfordenovoassemblyusingABySStoproduceconYgsConYgscanbe:AlignedtoexisYngreferencestoidenYfyvariantsinnewstrainsAnnotatedtoidenYfyputaYvegenes

ExtensiontoafulldraoreferencewillrequireaddiYonalsequencingtobuildscaffolds:

Mate-pairs: large insert, long reads to extend assembly 10X Chromium: Phased genomes, localized assemblies

Oxford nanopore: high throughput long reads Pacbio: consensus long read with lower error rate

Single-cellsequencing

WGS and RNA sequencing from individual cells allows for single cell resolution of copy number and expression

Cell populations treated under different conditions can be examined separately

Epigenomics

Post-transcripYonalmodificaYoncannotbedetectedthroughgenomeandtranscritomesequencing

EffortssuchastheInternaYonalHumanEpigenomeConsorYumhaveprovidedcomprehensivedatasetsforcomparisonandinterpretaYonofepigenomicdata

ChIPseqandbisulphitesequencing(array,wholegenomeorcapture)areusedtostudyhistonemodificaYonandDNAmethylaYon

Examplesofanalysis:IdenYfygenesandpathwaysthatareepigeneYcallymodified,correlatedChIPdatawithexpressionandmutaYonaldata,clustersamplesbyDNAmethylaYonprofile

ImmunogenomicsTCR/BCRsequencingHLAtypingAnalysisfromWGSandWTSsequencing:TandBcellrepertoireHLAtypingCelltypeabundanceNeoanYgenpredicYon

HowmuchdiskdoIneed

Data* Typicalfilesize

30Xgenome 50G

Full-depthtranscriptome 15G

miRNA 500M

1laneHiseq2500 50G

1laneHiseqX 65G

Variantfiles 10-100M

*human data, bam/fastq.gz/raw assembly data are similar in size

QuesYons

Aboutthistalk:yma@bcgsc.caAboutsequencingandbioinformaYcsattheGSC:dmiller@bcgsc.ca

AcknowledgementsImagesandresultsRichardCorbefReanneBowlbyDeniseBrooksEmiliaLimCaralynReisleEricChuahSteveBilobramWestgridJanaMakar

GSCMarcoMarra,DirectorStevenJones,DirectorGSCGroupleadersintheroomAndyMungall,BiospecimenandLibraryCoresRichardMoore,SequencingCoreRobinCoope,EngineeringDianeMiller,ProjectsMyawesomebioinforma)csteamRichard,Nina,Eric,Kane,Gina,Irene,Dorothy,Correy,Dean,Athena,Dan,Young,Brenna,Laura,Tina,Karen,Marc,Tori,Yisu,Patrick,Mya,Darryl,Nav,James,Brandon,Karen,Morgan,An,Diana,Johnson,Denise,Shirin,Marcus,Amir,Cara,DusYn,Caleb,Daniel,Simon,Steve,Reanne,Wei,Sara,StuartLab,Systems,Qualitysystems,Projects,Adminandallotherteams