CBC Data Therapy - University of Connecticut

CBCDataTherapyMetagenomicsDiscussion

GeneralWorkflow

Institute for Systems Genomics: Computational Biology Core

bioinformatics.uconn.edu

Microbial sample

Generate “Meta-omic” data

Process data (QC,

etc.)Analysis

MarkerGenes



Extract DNA

Amplify with

targeted primers

Filter errors, build

clusters

Diversity analysis

Metagenomics



Extract DNA

Sequence random

fragments

QC, assemble, annotate

Diversity, function analysis

Metatranscriptomics



Extract RNA,

subtract rRNA

SequencecDNA QC

Geneexpression,function

Sequencing



Sanger Ion Torrent Roche 454

Illumina *Seq Pacific Biosciences Nanopore

Resources(16S)



SILVA: Quast et al. NAR (2015)

rrnDB: Stoddard et al. NAR (2016)

RDP II: Cole et al. NAR (2013)

Resources(Genomes)



PATRIC (host and bacterial)GenBank Genomes

GOLD (JGI metagenomes) Ensembl Genomes

Resources(Metagenomes)



EBI metagenomicsMG-RAST

HMP DACC

Resources(Function)



KEGG

UniProtKB

CARD

Gene Ontology

GeneralChallenges/Considerations



• Sequencing errors• Error rates, error type (PacBio: 10% random, Illumina – 0.1% substitution)

• Chimeras• Amplification artifacts, cloning of restriction fragments

• 16S: different V regions give different results• Different sequencing platforms / sampling conditions ALSO give

different results• Workflow complexity / plethora of tools

GeneralChallenges/Considerations



•Strain-level diversity in metagenomes will often be missed by amplicon (esp. short-read) and shotgun approaches• This may be especially important between samples

•Taxonomy•Database predictions (RDP)

•Functional Annotation• Coverage versus accuracy

MarkerGenes



• EukaryoticOrganisms(protists,fungi)• 18S(http://www.arb-silva.de)• ITS(http://www.mothur.org/wiki/UNITE_ITS_database)

• Bacteria• CPN60(http://www.cpndb.ca/cpnDB/home.php)• ITS(Martiny,Env Micro2009)• RecA gene

• Viruses• Gp23forT4-likebacteriophage• RdRp forpicornaviruses

Fasterevolvingmarkersusedforstrain-leveldifferentiation

MarkerGenes



• Focusoncontaminationreductionduringpreparation• 16SrRNA contains9hypervariableregions(V1-V9)• V4waschosenbecauseofitssize(suitableforIllumina150bppaired-endsequencing)andphylogeneticresolution

• DifferentVregionshave differentphylogeneticresolutions– givingrisetoslightlydifferentcommunitycompositionresults

• Sequencing:• MiSeq capacityallowsmultiplesamplestobecombinedintoasinglerun• Numberofreadsneededtodifferentiatesamplesdependsonthenatureofthestudies• UniqueDNAbarcodescanbeincorporatedintoyourampliconstodifferentiatesamples

MarkerGenes



• QIIME(http://qiime.org)

• Mothur (http://www.mothur.org)

Bioinformatics



OverallBioinformaticsWorkflow

Sequencedata(fastq)

Metadata aboutsamples(mapping

file)

1)Preprocessing:removeprimers,

demultiplex,qualityfilter,decontamination

2)OTUPicking /RepresentativeSequences

5)SequenceAlignment

3)TaxonomicAssignment

4)BuildOTUTable(BIOMfile)

6)PhylogeneticAnalysis

PhylogeneticTreeOTUTable

7)DownstreamanalysisandVisualization

– knowledgediscovery

ProcessedData

Inputs

Outputs

QIIMEversusMOTHUR



QIIME Mothur

Apythoninterfaceto gluetogethermanyprograms Singleprogramwithminimalexternaldependency

Wrappersforexistingprograms Reimplementationofpopular algorithms

Large numberofdependencies/VMavailable Easy toinstallandsetup;workbestonsinglemulti-coreserverwithlotsofmemory

Morescalable Lessscalable

Steeperlearningcurvebutmoreflexibleworkflowifyou canwriteyourownscripts

Easytolearnandworksthebestwithbuilt-intools

http://www.ncbi.nlm.nih.gov/pubmed/24060131 http://www.mothur.org/wiki/MiSeq_SOP

Metagenomics



• Goal:Identifytherelativeabundanceofdifferentmicrobesinasamplegivenusingmetagenomics• Problems:

• Readsareallmixedtogether• Readscanbeshort(~100bp)• Lateralgenetransfer

• Twobroadapproaches1. BinningBased2. MarkerBased

Metagenomics



• Attemptsto“bin”readsintothegenomefromwhichtheyoriginated• Composition-based

• UsesGCcompositionork-mers (e.g.NaïveBayesClassifier)• Generallynotverypreciseandnotrecommended

• Sequence-based• ComparereadstolargereferencedatabaseusingBLAST(orsomeothersimilaritysearchmethod)• Readsareassignedbasedon“Best-hit”or“LowestCommonAncestor”approach

LCA



• UseallBLASThitsaboveathresholdandassigntaxonomyatthelowestlevelinthetreewhichcoversthesetaxa.

• NotableExamples:• MEGAN:http://ab.inf.uni-tuebingen.de/software/megan5/

• Oneofthefirstmetagenomic tools• Doesfunctionalprofilingtoo!

• MG-RAST:https://metagenomics.anl.gov/• Web-basedpipeline(mightneedtowaitawhileforresults)

• Kraken:https://ccb.jhu.edu/software/kraken/• Fastestbinningapproachtodateandveryaccurate.• Largecomputingrequirements(e.g.>128GBRAM)

Metagenomic Assembly



• “MetaSPAdes showedtheoverallbestassemblysizestatisticswhilealsocapturingarelativelylargefractionoftheexpecteddiversity.Theusageofthistoolisrelativelysimpleandconvenient,beingbasicallyidenticaltothatofSPAdes,andlargelyflexibleregardingtheformatoftheinputdata.Adrawbackmaybethereducedsensitivityformicrodiversity.However,forthemajorityofmetagenomeresearchquestions,accurateandrepresentativeconsensusgenomesofspeciesshouldbemorethansufficient. ”

CBC Data Therapy - University of Connecticut

Documents

Transcript of CBC Data Therapy - University of Connecticut