Whole genome shotgun sequencing DeNovo Assembly...1. Reference guided sequence assembly: map reads to a reference sequence 2. De-novo sequence assembly: determine overlap between sequence

WholegenomeshotgunsequencingDeNovoAssembly

2

StrategiestosequencelongDNAmolecules:ShotgunSequencing

TemplateDNA

1. RandomlybreaktemplateDNAintopieces

3


TemplateDNA

1. RandomlybreaktemplateDNAintopieces

4


1. RandomlybreaktemplateDNAintopieces2. Addadaptersofknownsequencetothefragmentendsandsizeselect

5

StrategiestosequencelongDNAmolecules:ShotgunSequencing Sometimesadaptersequencesremain!

1. RandomlybreaktemplateDNAintopieces2. Addadaptersofknownsequencetothefragmentends3. Sequence(typically)theendsofthefragments

Identifyingthesesequencesissimplewhenweignorethecomplexityofthesearch

Theproblemis,whatsequence(s)arewelookingfor?

6

StrategiestosequencelongDNAmolecules:ShotgunSequencing Sometimesadaptersequencesremain!

1. RandomlybreaktemplateDNAintopieces2. Addadaptersofknownsequencetothefragmentends3. Sequence(typically)theendsofthefragments

Identifyingthesesequencesissimplewhenweignorethecomplexityofthesearch

Clipadaptersequences

7


ReadPair1ReadPair2ReadPair3ReadPair4ReadPair5ReadPair6ReadPair7ReadPair8ReadPair9ReadPair10ReadPair11

1. RandomlybreaktemplateDNAintopieces2. Addadaptersofknownsequencetothefragmentends3. Sequence(typically)theendsofthefragments4. Identifyandremoveadapterpartfromthedeterminedsequences5. Reconstructtemplatesequencefromthesequencereads

Reconstructtemplate

8

1. RandomlybreaktemplateDNAintopieces2. Addadaptersofknownsequencetothefragmentends3. Sequence(typically)theendsofthefragments4. Removeadapterpartfromthedeterminedsequences5. Reconstructtemplatesequencefromthesequencereads

1. Referenceguidedsequenceassembly:mapreadstoareferencesequence2. De-novosequenceassembly:determineoverlapbetweensequencereadsandassemble

overlappingsequencesintocontigs.

Contig1 Contig2 Contig3

collapse

collapse

collapse

StrategiestosequencelongDNAmolecules:Shotgunsequencingandde-novoassemblyofthesequencereads

9

ReadPair11ReadPair2

ReadPair5


Scaffold1NN

1. RandomlybreaktemplateDNAintopieces2. Addadaptersofknownsequencetothefragmentends3. Sequence(typically)theendsofthefragments4. Removeadapterpartfromthedeterminedsequences5. Reconstructtemplatesequencefromthesequencereads

1. Referenceguidedsequenceassembly:mapreadstoareferencesequence2. De-novosequenceassembly:determineoverlapbetweensequencereadsandassemble

overlappingsequencesintocontigs.Matepairinformationcanthenbeusedtobuildsuper-contigs(scaffolds)fromphysicallynon-overlappingcontigs.

Scaffold2NNNN NN

StrategiestosequencelongDNAmolecules:Shotgunsequencingandde-novoassemblyofthesequencereads

10

1. Coverage:TheaveragenumberofreadscoveringapositioninthesequencedtemplateDNA.Lengthofgenomicsegment:LNumberofreads:nCoverage C=nl/LLengthofeachread:l

Howmuchcoverageisenough?Lander-Watermanmodel: Assuminguniformdistributionofreads,C=10resultsin1gappedregionper1,000,000nucleotides->Thisisnomorethanacruderuleofthumbandgreatlydependsonreadlength,repeatcompositionofthetemplateDNA,etc.


collapse

collapse

collapse

StrategiestosequencelongDNAmolecules:Shotgunsequencingandde-novoassemblyofthesequencereadsSomesummarystatisticstodescribeassemblies

Thehigherthecoveragethebetter(providedunlimitedcomputationalresources)!Themoreuniformthecoveragedistributionthebetter!

11

2. N50-size:Morethan50%ofthebasesinyourassemblyresideincontigswithatleastthesizedeterminedbytheN50value.NOTE:YoucanofcoursespecifyanyotherN-value.


collapse

collapse

collapse


WhatnowtellsustheN50sizeexactly?Isitaqualitymeasureaspeoplefrequentlyuseit?

WhendoesitmakesensetomentiontheN50size(justconsiderRNAseqassemblies)?

12

3. Contiglengthdistribution


collapse

collapse

collapse


Note:Allthesestatisticscanbeusedforscaffoldsaswell!

Literatureonde-novoassemblies

• J.R.Milleretal.Genomics95(2010)315–327• P.Compeauetal.NatureBiotechnology29(2011)987-991

• ZerbinoandBirney.GenomeRes18(2008)821-829Velvet

DeNovoAssembly

Theassemblyproblem…

JLFB

JLFB

Overlap-Layout-ApproachAssemblers:ARACHNE,PHRAP,CAP,TIGR,CELERA,MIRA

Overlap:findpotentiallyoverlappingreads

Layout:mergereadsintocontigsandcontigsintosupercontigs

Consensus:derivetheDNAsequenceandcorrectreaderrors ..ACGATTACAATAGGTT..

Overlap

• Findthebestmatchbetweenthesuffixofonereadandtheprefixofanother

• Duetosequencingerrors,needtousedynamicprogrammingtofindtheoptimaloverlapalignment

• Applyafiltertoremovepairsoffragmentsthatdonotshareasignificantlylongcommonsubstring

Differentapproachestothesequenceassemblyproblem

modfiedfromCompeauetal.(2011)NatureBiotechnology29(11)

Overlapbasedassembly➢ readidentityismaintained➢ intuitive➢ Readscanbeorganisedinanoverlapgraph➢ Graphcomplexityincreaseswithcoverage,

thusreadredundancyinflatesthegraph

Word-basedapproaches➢ readidentityis(temporarily)

lost…➢ ReadsareorganisedindeBruijn

graphs➢ Graphcomplexitydependson

wordsize➢ Graphcomplexityis(byand

large)independentfromcoverage,readredundancyisnaturallyhandled

➢ repeatsarerepresentedonlyonceinthegraphwithexplicitlinkstothedifferentstartandendpoints

De-NovoSequenceAssembly:  CAP3

18

ShotgunSequenceReads

1234567

De-NovoSequenceAssembly:  CAP3

19


1. Concatenatesequencesintoacombinedsequence.Readsareseparatedbyaseparationcharacter.

1234567

1 2 3 4 5 6 7

De-NovoSequenceAssembly(CAP3)  Searchforlocalalignments

20



2. Computehighscoringchainsofsegmentsbetweeneachreadandthecombinedsequenceusinglocalalignmentsearchtools.Identifycandidatepairs.Everypairiscountedonlyonce.

234567

1 2 3 4 5 6 7


21




1

234567

1 2 3 4 5 6 7


22




1

234567

1 2 3 4 5 6 7

Removethetrivialsolution(alignmentagainstitself)


23




234567

1 2 3 4 5 6 7

Removethetrivialsolution(alignmentagainstitself)


24




1

234567

1 2 3 4 5 6 7

1


25




1

234567

1 2 3 4 5 6 7

1

1

3 5

1Candidatepairsforread1:


26




34567

1

3 5


2

1 2 3 4 5 6 7

2


27




234567

1 2 3 4 5 6 7

2

1

3 5


6


2

3


28




4567

1 2 3 4 5 6 7

1

3 5


6


Candidatepairsforread3:

3

6

3

3

7

3

2

3


29




4567

1 2 3 4 5 6 7

1

3 5


6


Candidatepairsforread3:6

3

7

3

Candidatepairsforread7:

.

.

.

2

3

De-NovoSequenceAssembly(CAP3)  Searchforlocalalignments:post-processing

30

1

3

5

1

6

2

6

3

7

3



3. Removepoorqualitysequenceends

4. Computeglobalalignmentforthehighqualitysequencepairstoverifyoverlaps.

2

3

4


31

1

3

5

1

6

2

6

3

7

3




4. Computeglobalalignmentforthehighqualitysequencepairstoverifyoverlaps.Evaluateaccordingtothefollowingcriteria:1. minimumlength2. minimumidentity3. minimumsimilarity4. numberofhigh-qualitymismatches

Removesequencepairsthatdonotmeetthethresholdsfor4.1to4.4

4

2

3


32

1

3

5

1

6

2

6

3

7

3






4

2

3


33

1

3

5

1

6

2

6

3

7






4

2

3

1

3

16

2

6

3

2

3

CAP3:ContigBuilding

5

4 7

34

5

16

2

6

3

CAP3:ContigBuilding

1

3

2

4 7

35

5

16

2

6

CAP3:ContigBuilding

1

3

2

4 7

36

5

6

CAP3:ContigBuilding

13

2

6

2

4 7

37

5

6

CAP3:ContigBuilding

13

26

2

1) Generateagenerallayoutusingtheoverlappingreadsfromthepair-wise

analysis(Greedyalgorithmindecreasingorderofoverlapscores).

2) Inasimpleview:Checkthelayoutforincompatibilities.

4 7

38

5

6

CAP3:ContigBuilding

13

26

2




1) sequenceread1and2areincompatiblesincetheycouldnotbe

aligned.

4 7

39

5

6

CAP3:ContigBuilding

13

6

2





aligned.

2) resolveincompatibility

3) checkfornewpossiblelayouts

4 7

40

5

6

CAP3:ContigBuilding

13 2





aligned.

2) resolveincompatibility

3) checkfornewpossiblelayouts

4 7

41

5

6

Furtherstepsingenomesequencing

13 2



2) Inasimpleview:Checkthelayoutforincompatibilities,remove

incompatiblereadsandalign.

3) Buildaconsensussequenceforeachcontigs.

4) Orderandorientcontigsifpossibleusingadditionalinformation,e.g.,

pairedendreads.

4 7

42

DeriveConsensusSequence

Derivemultiplealignmentfrompairwisereadalignments

TAGATTACACAGATTACTGA TTGATGGCGTAA CTATAGATTACACAGATTACTGACTTGATGGCGTAAACTATAG TTACACAGATTATTGACTTCATGGCGTAA CTATAGATTACACAGATTACTGACTTGATGGCGTAA CTATAGATTACACAGATTACTGACTTGATGGGGTAA CTA

TAGATTACACAGATTACTGACTTGATGGCGTAA CTA

Derive each consensus base by weighted voting

Errorcorrectionbyweightedvoting

• Correcterrorsusingmultiplealignment

TAGATTACACAGATTACTGATAGATTACACAGATTACTGATAG TTACACAGATTATTGATAGATTACACAGATTACTGATAGATTACACAGATTACTGA

C: 20C: 35T: 30C: 35C: 40

C: 20C: 35C: 0C: 35C: 40

• Scorealignments• Acceptalignmentswithgoodscores

A: 15A: 25A: 40A: 25-

A: 15A: 25A: 40A: 25A: 0

Consensus(cont’d)• Aconsensussequenceisderivedfromaprofileofthe

assembledfragments

• Asufficientnumberofreadsisrequiredtoensureastatisticallysignificantconsensus

• Readingerrorsarecorrected

SequenceassemblywithDeBruijngraphs

Sequenceassemblycanbeabstractedtotheshortestsuperstringproblem  

(NicolasdeBruijn1946)

Problem:findtheshortest(circular)superstringthatcontainsallpossiblesubstringsoflengthKoveragivenalphabet.

forK=4andatwoletteralphabetA={0,1}wehave16differentwords:

0000,0001,0010,0100,1000,0011,0110,1100,1001,1010,0101,0111,1011,1101,1110,1111

Sequenceassemblycanbeabstractedtotheshortestsuperstringproblem  

(NicolasdeBruijn1946)

Problem:findtheshortest(circular)superstringthatcontainsallpossiblesubstringsoflengthKoveragivenalphabet.

forK=4andatwoletteralphabetA={0,1}wehave16differentwords:

0000,0001,0010,0100,1000,0011,0110,1100,1001,1010,0101,0111,1011,1101,1110,1111

Tosolvethisproblem,deBruijnborrowedfromEulerwhosolved1735the‘Königsberg’problem,i.e.thequestionwhetheritispossibletovisiteachislandbycrossingeachbridgeexactlyonce(Euleriancycle)

EulerianCycleProblem

• Findacyclethatvisitseveryedgeexactlyonce(Lineartime)

• AnEulerianCycleexistsifthenumberof‘outgoing’edgesforanodeequalsthenumberof‘incoming’edges*.

• Thegraphmayhave2nodeswithanoddnumberofedgesconnectedtoit.InthiscaseanEulerianpathratherthananEuleriancyclecanbefound.

*ProofbyC.Hierholzer

deBruijnsolvedtheproblembyrepresentingK-1mersasnodesandKmersasedgesinadirectedgraph.

Bydoingso,herelatedtheproblemoffindingashortestcommonsuperstringtothealreadysolvedproblemoffindinganEuleriancyclein

agraph.

001 011

100 110

000 010 101 111

0011

0010

0101

10111010

1101

0111

1111 1110

1100

10010001

10000000

0110

0100

I

II

III

IV

V

VI

VII

VIII

IXX

XI

XIIXIII

XIV

XV

XVI

001 011

100 110

000 010 101 111

0011

0010

0101

10111010

1101

0111

1111 1110

1100

10010001

10000000

0110

0100

I

II

III

IV

V

VI

VII

VIII

IXX

XI

XIIXIII

XIV

XV

XVI

Passingthroughtheedgesbyfollowingtheromannumbersreconstructsthesuperstringusingeachwordexactlyonce!

I:0000,II:0001,III:0011;IV:0110;V:1100;VI:1001;VII:0010;VIII:0101;IX:1011;X:0111;XI:1111;XII:1110;XIII:1101;XIV:1010;XV:0100;XVI:1000

0000110010111101

AdvancingtoDNAsequenceassemblyisstraightforward

modfiedfromCompeauetal.(2011)NatureBiotechnology29(11)

Classicaloverlapassembly(readidentityismaintained)

Kmerapproaches(readidentityislost)

DeBruijnGraphExample Shredreadsintok-mers(k=3)

G G A C T A A

G G A

G A C

A C T

C T A

T A A

G A C C A A A

G A C

A C C

C C A

C A A

A A A

53

Read1 Read2

GG(1x)

GA(1x)

AC(1x)

CT(1x)

TA(1x)

AA(1x)

GGA GAC ACT CTA TAA

GA(1x)

AC(1x)

CC(1x)

CA(1x)

AA(1x)

AA(1x)

GAC ACC CCA CAA AAA

DeBruijnGraphExample Mergeverticeslabeledbyidentical(k-1)-mers

Read1:

Read2:

54

ResultingGraph:GG(1x

GA(2x)

AC(2x)

CT(1x)

TA(1x)

AA(2x)

CC(1x)

CA(1x)

AA(1x)

GG(1x)

GA(1x)

AC(1x)

CT(1x)

TA(1x)

AA(1x)

GA(1x)

AC(1x)

CC(1x)

CA(1x)

AA(1x

AA(1x)

AnotherExample  Constructthegraph(k=5)

AGAT(8x)

ATCC(7x)

TCCG(7x)

CCGA(7x)

CGAT(6x)

GATG(5x)

ATGA(8x)

TGAG(9x)

GATC(8x)

AAGT(3x)

AGTC(7x)

GTCG(9x)

TCGA(10x)

GGCT(11x)

TAGA(16x)

AGAG(9x)

GAGA(12x)

GACA(8x)

ACAA(5x)

GCTT(8x)

GCTC(2x)

CTTT(8x)

CTCT(1x)

TTTA(8x)

TCTA(2x)

TTAG(12x)

CTAG(2x)

AGAC(9x)

CGAG(8x)

CGAC(1x)

GAGG(16x)

GACG(1x)

AGGC(16x)

ACGC(1x)

55

Abranchingvertexiscausedbyeitherarepeatintheoriginalsequenceorasequencingerror

Sequencingerrorsaretypicallydetectedbyacoveragecutoffthreshold

Condenseunbranchedrunsinthegraph

AAGTCGA

TAGAGCTTTAG

GCTCTAG

GAGACAA

CGAG

CGACGC

GAGGCT

AGATCCGATGAG

56

AGAG

Correctsequencingerrorsusingacoveragethreshold

AAGTCGA

TAGAGCTTTAG

GAGACAA

CGAG

GAGGCT

AGATCCGATGAG

57

AGAG

Afterrecondensation

AAGTCGAG GAGACAAGAGGCTTTAGA

AGATCCGATGAG

58

AGAG

Source:SerafimBatzoglou

Anynon-branchingpathinthisgraphcorrespondstoacontigintheoriginalsequence.

Takingtheriskoffollowingarbitrarybranchingpathsmaycreatechimericspecies

BasicconceptsofdeBruijngraphbasedassemblers

• ThesequenceistreatedasaconsecutivestringofwordsoflengthK

• Sequencereadsarenolongerconsideredtorepresentaconsecutivestringofnucleotides.Thusreadlengthaswellasreadoverlapbecome,inprinciple,irrelevant.

• SequencereadsareonlyusedtoidentifywordsoflengthKoccurringinthesequence.

• Givenperfectdata–error-freeK-mersprovidingfullcoverageandspanningeveryrepeat–theK-mergraphwouldbeadeBruijngraphanditwouldcontainanEulerianpath,thatis,apaththattraverseseachedgeexactlyonce.

Themagic‘Kmer’givesmostusersofgraphbasedassemblyalgorithmsaveryhardtimeastheyhavetodecideonthesizeofK.

TogiveaninformedstatementweneedtomakesuretounderstandwhatKshouldrepresentandwhatthealgorithmicrequirementsofdeBruijngraphassemblersare

• Kmustrepresentawordthatoccursonlyonceinthesequencethatshouldbeassembled.Thus,Kmustbesufficientlylarge.

• deBruijngraphbasedassemblersassumethateachwordoflengthKoccurringinthegenomeisalsorepresentedinthegraph.AsKmersarecollectedfromafinitesetofsequencereads,Kmustnotbetoolarge.

• consideraDNAwordofK=2,howoftendoesitonaverageoccurinastringof16bp?

• HowaboutawordofK=25*

DefaultforTrinity(Grabherretal.2011)

Takehomemessage:IfKisonlysufficientlylargethechanceforanyKmertooccurmorethanonceina

(repeat-free)genomeapproaches0.

WhynotusingsimplythereadlengthasK?

WhyKmustnotbetoolarge

AGACTAGAGAATTGCGATAG

Asequenceoflength20contains11differentwordsoflength10!

Now,considerthesequenceisspannedby2readsoflength13:

AGACTAGAGAATTGCGATAGAGACTAGAGAATT

AGAATTGCGATAG

T:R1:R2:

Itiseasytoseethatnotall11wordsoflength10canbereconstructedwiththetworeads.ThisviolatesthekeyassumptionofthedeBruijngraphsItisalsoeasytoseethatreducingKamelioratestheproblemandeventuallygetsridofit(justconsiderK=1…)

k = 151 N50: 57 kbp

Reference: 40%

Assembly parameter optimization using maximization of average contig length (N50) as objective can

preclude an entire genome from being assembled

k = 51 N50: 22 kbp

Reference: 99%

Fungal kmers

Algal kmers & seq errors

Fungal kmersAlgal kmers

seq errors

Documents

Whole genome shotgun sequencing DeNovo Assembly...1. Reference guided sequence assembly: map reads to a reference sequence 2. De-novo sequence assembly: determine overlap between sequence