Evaluation of Compatibility of 16S rRNA V3V4 and V4 ...

The 16S V3V4 and V4 hypervariable regions are widely selected for human microbiota profiling, but the fragments amplified from different regions ... Skiptomaincontent NewResults EvaluationofCompatibilityof16SrRNAV3V4andV4AmpliconLibrariesforClinicalMicrobiomeProfiling ViewORCIDProfilePo-YuLiu,Wei-KaiWu,Chieh-ChangChen,SuraphanPanyod,Lee-YanSheen,Ming-ShiangWu doi:https://doi.org/10.1101/2020.08.18.256818 Po-YuLiuaDepartmentofInternalMedicine,NationalTaiwanUniversityCollegeofMedicine,Taipei,TaiwanFindthisauthoronGoogleScholarFindthisauthoronPubMedSearchforthisauthoronthissiteORCIDrecordforPo-YuLiuWei-KaiWubDepartmentofInternalMedicine,NationalTaiwanUniversityHospitalBei-HuBranch,Taipei,TaiwancInstituteofFoodScienceandTechnology,NationalTaiwanUniversity,Taipei,TaiwanFindthisauthoronGoogleScholarFindthisauthoronPubMedSearchforthisauthoronthissiteChieh-ChangChenaDepartmentofInternalMedicine,NationalTaiwanUniversityCollegeofMedicine,Taipei,TaiwandDivisionofGastroenterologyandHepatology,DepartmentofInternalMedicine,NationalTaiwanUniversityHospital,Taipei,TaiwaneGraduateInstituteofClinicalMedicine,NationalTaiwanUniversityCollegeofMedicine,Taipei,TaiwanFindthisauthoronGoogleScholarFindthisauthoronPubMedSearchforthisauthoronthissiteSuraphanPanyodcInstituteofFoodScienceandTechnology,NationalTaiwanUniversity,Taipei,TaiwanFindthisauthoronGoogleScholarFindthisauthoronPubMedSearchforthisauthoronthissiteLee-YanSheencInstituteofFoodScienceandTechnology,NationalTaiwanUniversity,Taipei,TaiwanFindthisauthoronGoogleScholarFindthisauthoronPubMedSearchforthisauthoronthissiteMing-ShiangWuaDepartmentofInternalMedicine,NationalTaiwanUniversityCollegeofMedicine,Taipei,TaiwandDivisionofGastroenterologyandHepatology,DepartmentofInternalMedicine,NationalTaiwanUniversityHospital,Taipei,TaiwanFindthisauthoronGoogleScholarFindthisauthoronPubMedSearchforthisauthoronthissiteForcorrespondence: [email protected] AbstractFullTextInfo/HistoryMetricsPreviewPDF ABSTRACTSequencingofthe16SrRNAgenebyIlluminanext-generationsequencingisbroadlyusedinmicrobiomestudies.Differenthypervariableregionsofthe16SrRNAgene,V3V4(amplifiedwithprimers341F–805R)orV4(V4O;primers515F–806R),areselected,dependingonthetargetedresolution.However,inpopulation-basedclinicalstudies,combiningV3V4andV4datafromdifferentstudiesforameta-analysisischallenging.Readsgeneratedbyshort-read(150-bp)high-throughputsequencingplatformsdonotfullyrecovertheV4regionread-length.Here,weevaluatedthecompatibilityof16SrRNAV3V4andV4ampliconsformicrobiomeprofiling.WecomparedtaxonomiccompositionsobtainedbytheanalysisofV3V4andV4amplicons,andV4fragmentstrimmedfromV3V4amplicons.WealsoevaluatedanalternativeV4region(V4N;primers519F–798R)designedforefficientstitchingwith150-bppaired-endsequencing.First,wesimulatedaglobalinvestigationofenvironmentalprokaryotesinsilico.ThisrevealedthatV4Oprimersrecoveredthehighestproportionoffragments(81.7%)andmostphyla,includingarchaea.Empiricalsequencingofstandard(mock)andhumanfecalsamplesrevealedbiasedpatternsofeachprimerthatweresimilartotheonesdeterminedbyinsilicosimulation.Further,forhumanfecalmicrobiomeprofiling,thebetween-samplevariancewasgreaterthanthesystematicbiasofeachprimer.TheuseoftrimmedV4fragmentsandsingle-endampliconsresultedinthesamesystematicbias.Inconclusion,paired-endV4Osequencingyieldedthemostaccuratedataforboth,simulationandmockcommunitysequencing;theV4OampliconswerecompatiblewithtrimmedV4sequencesformicrobiomeprofiling.IMPORTANCENext-generationsequencingofthe16SrRNAgeneisacommonlyusedapproachforclinicalmicrobiomestudies.Differentampliconsofthe16SrRNAhypervariableregionsareusedindifferentstudies,whichcreatesincompatiblesequencefeatureswhencomparingandintegratingdataamongstudiesbyusing16Sdenoisingpipelines.Herewecomparedthetypeofdataandcoverageobtainedwhendifferent16SrRNAampliconswereanalyzed.InsilicoandempiricalanalysesofthehumanfecalmicrobiomerevealedthattheV3V4ampliconsarecompatiblewithV4ampliconsaftertrimminguptothesameregion.Theseobservationsdemonstratethatreconcilingthecompatibilityofclinicalmicrobiomedatafromdifferentstudiesimprovenotonlythesamplesizebutalsotheconfidenceofthehypothesistested.INTRODUCTIONThehumanmicrobiomeaffectshumanhealthinnumerousways(1-7).Notably,thegutmicrobiomemediateshostmetabolicandphysiologicalstatus,suchasdigestion,immuneresponse,neurontransmission,andcirculation(8).Clinicalresearchofthemicrobiomeisrapidlyadvancing,propelledbysequencingofthe16SrRNAgene(abbreviated“16S”)usingthenext-generationsequencing(NGS)technology.Althoughlong-readsequencingtechnologieshavematuredinrecentyears,thewell-developedanalysispipelinesformassivemicrobiometaxonomicprofiling(e.g.,QIIME2)aremainlybasedontheIlluminasequencersystems(9).Accurateevaluationofthemicrobiotaheavilydependsontheprimersused(10,11).Further,lower-leveltaxonomicresolutionbiascanarisewhennon-representativeregionsareamplified(12).The16SV3V4(primers341F–805R)andV4(primers515F–806R)hypervariableregionsaremostfrequentlyusedforhumanmicrobiomeprofiling(6,13).WhenanampliconlibraryispreparedusingNexteraXTtwo-steppolymerasechainreaction(PCR),theexpectedinsertsizesfromtheseregionsare465bpand291bp,respectively.(14,15).TheV3V4andV4ampliconsarefullyrecoveredontheIlluminaMiSeqplatform,whichcanbeusedtosequenceupto600nucleotidesfrombothendsofanamplicon[(300bp)×2].WhentheNationalInstituteofHealthHumanMicrobiomeProject(HMP)startedin2008,investigationsofthehumanmicrobiotausingRoche/454pyrosequencingfocusedonthe16SV3V5region(primers357F–926R)(16).However,theV3V4regionbecamethemainstreamamplicontargetinmicrobiotastudiessinceIlluminareleasedarecommendedlibrarypreparationprotocolforsequencingontheMiSeqplatform(15).Althoughtheoutputsofbothsequencingapproacheswerecomparable(17),theIlluminaplatformgeneratedmuchmorereadsthanpyrosequencing.AfterHMP,in2010,theEarthMicrobiomeProject(EMP)wasinitiated,asaglobalinvestigationofenvironmentalandhost-associatedmicrobiomes(18,19).The16SV4region(primers515F–806R)wasamplifiedforsequencingintheEMPprojects(18),includinghuman-associatedmicrobiomestudies(20)andtheAmericanGutProject(21).ThisallowedforamorerepresentativeprokaryoticprofilingsincetheV4universalprimerpaireffectivelycapturesbothbacterialandarchaeal16Ssequences.However,theEMPV4librarieswereconstructedusingcustomsequencingprimersaccordingtoCaporasoetal.(22),andgeneratedby150-bppaired-end(PE)sequencingwithadditional14cyclesforbarcodetaggingonMiSeqorHiSeq.TheampliconinsertsizeofthecustomV4libraryisexpectedtobe252bp,whilethetwo-stepPCRmethodgenerates291bpfromthesametargetedregion(14).ThismeansthattheNexteraXTtwo-stepPCRmethodisonlysuitableforplatformswithanover150-bpPEsequencingcapacity(250-bpPEand300-bpPE).Nonetheless,theV4regionisanalyzedregardlessofwhichprotocolisfollowed.Datafromdifferentprotocolsaresupposedtobeideallyintegratedampliconsequencevariants(ASVs)formeta-analysismicrobiotaprofilingbyusingdenoisingalgorithms(23)[e.g.,DADA2(24),Deblur(25),andUNOISE3(26,27)]intheQIIME2pipeline(28).ThesequencingcostpermegabasehasexceededMoore’sLaw,whichdescribesatrendindoublingofcomputingpowerthatconceivestheimprovementofDNAsequencingcapacityyearlyandevenovertheexponentialtrend(29).TheIlluminasequencingplatformsgeneratebetweenseveraltenstohundredsofmillionsofreads,enablingdeepprofilingofalargenumberofsamplesduringasinglePErunatafractionofthecostofastudy.Theoutputofmorethan100,000readspersampleissuggestedandsufficientformicrobiotainvestigations(15).Reducedsequencingrunsarepreferredinlargesample-sizestudies,especiallyclinicalcohortstudies,toavoidbatcheffects.Higherthroughputsequencers,suchasNextSeqandHiSeq,whicharecommoninacademiccorelaboratories,generatereadswithalowbatcheffectandatalowcostpersamplewithasinglerun(30).However,theconfidenceforstitchingbothendstorecoverthefullV4regionafterqualitytrimmingofthusgeneratedPEsequences[(150bp)×2,themaximumreadlengthofNextSeq]islow.Hence,itisdifficultforthehigherthroughputplatformstomeetthedemandsforcost–benefitanddatayields.Inresponsetothechangesofampliconsequencingmethodsintheclinicalgutmicrobiotaresearchandconsideringthecost–benefitofsequencing,weherecomparedseveralpairsofsequencingandanalysisapproaches.WefirstconductedaninsilicoPCRsimulationofaglobalinvestigationofenvironmentalprokaryotesbycapturingthe16SV3V4andV4regions,aswellastheV4fragmentstrimmedfromV3V4fromtheSILVA132ribosomalRNANR99database(DB).Wetestedprimerandanalysisbiasbyanalyzingamockmicrobialcommunity.WethensequencedhumanfecalsamplesandanalyzedthedatausingtheQIIME2pipelinewiththeDADA2plugin,whichiscurrentlythemostaccuratesampleinferencemethod(denoising).Wethenevaluatedthecompatibility(includingthetaxonomicabundanceconsistencyandcoverage)oftheampliconsofdifferent16Shypervariableregions.Theanalysisrevealedprimer-andanalysismethod-associatedbias.Basedonthefindings,weproposeoptimizedanalyticaloptionsforclinicalpopulation-basedandmeta-analysisstudies.RESULTSCostevaluationof16SampliconsequencingusingtheIlluminaMiSeqandNextSeqplatformsNGSsequencersareversatileplatformsforsequencing-basedstudies.Thecost–benefitratio(outputquantityandquality)ofsequencinghastobeconsideredwhenchoosingasuitableplatform.Focusingonthe16Sampliconsequencing,wecomparedthecosteffectivenessoftheIlluminamidtomid–highthroughputplatforms(MiSeqandNextSeqplatforms),whichthemajorityofacademiccorelaboratoriesarecommonlyequippedwith(Table1).Viewthistable:ViewinlineViewpopupDownloadpowerpointTABLE1Summaryoftheestimated16SampliconsequencingthroughputsandcostsforMiSeqandNextSeqplatformsThemaximalflowcellconfigurationsofMiSeqandNextSeqare600(300-bpPE)and300(150-bpPE)cycles,respectively.BasedontheIlluminaNexteraXTlibrarypreparationmethodfor16Samplicons,theV3V4andV4regionsarefullyrecoveredusingtheMiSeq600-cyclereagentkits;ontheotherhand,onlytheV4regioncanbein“theory”recoveredusingtheNextSeq300-cyclereagentkits.Fixingtheoutputat100KPEreadspersample,theMiSeqcanprocess175samplesperrun(96samplesperrunastheregularconfiguration),usingthedefaultlibrarypreparationmethod(NexteraXT).Themid-outputandhigh-outputflowcellsoftheNextSeqcanprocessupto384samplesperrun(allindexpairs),accountingfor39.4%and12.8%readsofasinglerun,respectively.The16SampliconsequencingcostsinUSdollarsare$15.2per100KreadspersamplefortheMiSeq,decreasingto$3.4and$2.8fortheNextSeqmid-outputandhigh-outputplatforms,accordingly.SincetheIlluminaNGSbasequalitydecreasestowardthe3’-endovertheread(Fig.1A),theoverlappingareaofthePEsequencesshouldbesufficientlylongtoallowtheassemblybyhigh-qualitybasepairs(Fig.1B).However,whentheV4ampliconsequencingisperformedviashortread-lengths(e.g.,150-bpPE),gapsexistbetweenthetwoampliconendsafterqualitytrimming(Fig.1C).Caporasoetal.(29)modifiedtheV4librarypreparationmethodtomakeitsuitableforusewith150-bpPEsequencing.Themodifiedmethodgenerates252-bpampliconsinsteadof291-bpampliconsobtainedwiththeNexteratwo-stepPCRapproach.However,althoughthemodifiedmethodincreasesboth,theV4ampliconassemblyefficacyandsequencingcapacity(upto975and2168samplesperrunontheNextSeqmid-outputandhigh-outputplatforms),itrequiresmanualalterationofthesequencingsoftwareconfiguration.Theabovecomparisonsrevealedthatitisnecessarytooptimizesequencingconfigurations(i.e.,libraryconstructionbyamplifyingsuitablehypervariableregionandsequencingwithlongenoughconfiguration)forimprovingthecost-benefitof16Sampliconsequencing.We,therefore,furtherproceededtotestiftheshorter(V4)ampliconsharborequivalentorbettertaxonomicprofilingcapacitiescomparedtothelonger(V3V4)amplicons.DownloadfigureOpeninnewtabFIG1NGSread-qualitypatterndiagram.(A)Illuminaread-qualitypatterns:fromthe5’-endto3’-end,thequalityperbasedecreasesandisusuallytrimmed(reddashedline)beforeassembly.(B)The300-bppaired-end(PE)read-qualitydistributionandampliconstitchingprocedures.The300-bpPEreadswithsufficientlylongoverlappingregionsarestitchedtogether.(C)The150-bpPEread-qualitydistributionandampliconstitchingprocedures.Onlyafewread-pairspassthequalityfilteringandassemblyatbothends.Most150-bpPEreadseitherfailtopassthequalitycontrolorfailtorecoverampliconinsertsthataretoolong.Comparisonoftaxonomicprofilingcapacitiesofthe16SV3V4andV4regionsbyaninsilicoPCRsimulationToevaluatetheprimerefficaciesof16SV3V4(primers341F–805R)andV4primerpairs(includingV4original,V4O,primers515F–806R;alternativeV4,V4N,primers519F–798R;tV4O,V4OregiontrimmedfromV3V4fragment;andtV4N,V4NregiontrimmedfromV3V4fragment),wesimulatedPCRcaptureofthetargetedfragmentsfromtheDB[SILVA16Sgenedatabase(NR13299%)](31,32).Thesimulationencompassedaninvestigationoftheprimer-dependencyofdetectedbacterialandarchaealprofilesbecausetheinsilicoPCRcapturedalltargetedsequencesfromtheglobalenvironment.TheDBcontains369,953representative16Ssequences(Table2).Ourtargetedapproachesextracted59.2%to81.7%ofsequencesfromtheDB.AlthoughtheV3V4primerscapturedthelongestfragments,theyextracted77.3%ofallDBsequences,whiletheV4Oprimersextracted81.7%ofallsequences.TheV4N-capturedsequencescoveredmostoftheV4Oregionbutwereshorterbyapproximately11bp.ThisreducedthecapturerateofV4Nto66.2%.ThetV4primersrecovered72.5%(tV4O)and59.2%(tV4N)ofsequencesfromtheDB,and93.8%(tV4O)and76.5%(tV4N)oftheV3V4sequences.Viewthistable:ViewinlineViewpopupDownloadpowerpointTABLE2CoverageratesofinsilicoPCRsimulationwiththeV3V4andV4primersBasedontheassignedtaxonomyanalysis,allprimerscapturedsimilarproportionsofthemajorphylaininsilicoPCR(Fig.2A).However,theoveralldifferencesfromtheDBcompositionrangedfrom6.44%to36.69%ofthephylaproportions.TheV4Oresultedintheleastdifferences(6.44%)fromtheDB,followedbyV3V4(15.23%)andV4N(30.0%).ThetrimmedV4approachesledtodatasetsthatdifferedby16.9%(tV4O)and36.69%(tV4N)fromtheDB.BothV4OandV4Napproachescapturedthemaximumarchaea(5.26%and3.40%,respectively;TableS1),whiletheV3V4approachcapturedthefewestarchaea(0.02%).ThetwotrimmedV4approaches,designedbasedonV3V4,didnotefficientlycapturethearchaea(0.02%fortV4Oand0.005%fortV4N).DownloadfigureOpeninnewtabFIG2TaxonomicprofilingviainsilicoPCRsimulation.(A)PhylumcompositionoftheSILVA16SdatabaseandextractedphylumcompositionforthefivetestedconditionsofinsilicoPCR.(B)Phylumtospeciestaxonomicaccuracies(againsttaxonomicassignmentintheSILVA16Sdatabase)forthefiveconditionsofinsilicoPCR.ThefiveconditionsforinsilicoPCRwereasfollows:V3V4(341F–805R),V4O(515F–806R),V4N(519F–798R),tV4O(trimmedV4OfromV3V4),andtV4N(trimmedV4NfromV3V4).Wethencomparedtheaccuracyoftheclassificationateachtaxonomiclevelbythedifferentprimerapproaches(Fig.2B).Over90%ofV4O-generatedsequenceswereassignedcorrecttaxonomy,followedbytheV3V4andtV4Osequences(approximately80%accuracyateachlevel).However,atmost70%and63%oftheV4NandtV4Nsequences,respectively,wereassignedthecorrecttaxonomy(summarizedinTable3).ThisindicatesthattheV4Oprimerwouldyieldthebesttaxonomicprofilesinaglobalmicrobialinvestigation.Viewthistable:ViewinlineViewpopupDownloadpowerpointTABLE3Summaryofsequencingefficacies(insilicoandempiricaldata),consideringthecoveragerate,taxonomicidentificationrate,andtaxonomicabundance,anddependingontheprimersandanalysisapproachesaMockcommunityprofilingbyusingsevendifferent16SampliconanalyticapproachesToempiricallyevaluatethecompatibilityofV3V4andV4primers,wesequencedandanalyzedamockmicrobialcommunity.Themockcommunityrepresentedeightpspeciesofbacteriaandtwospeciesofyeasts(seeMaterialsandMethods)andisanartificialsynthesizedmicrobialcommunitythatservesasaquantitativestandard.Weusedsevenanalyticalapproaches,allinconjunctionwiththeQIIME2DADA2denoisingpipeline(9,24)(Fig.3),i.e.,weanalyzedPEV3V4amplicons(V3V4),PEV4Oamplicons(V4O),PEV4Namplicons(V4N),single-endV4Oamplicons(V4OSE),single-endV4Namplicons(V4NSE),andV4ampliconstrimmedfromV3V4(trimmedV4O,tV4O;andtrimmedV4N,tV4N).Wealsosequencedhumanfecalsamplesfrom10volunteersusingthesameprotocolsutilizedtoevaluatetheprimercompatibilitiesinrealtargetedsamples.DownloadfigureOpeninnewtabFIG3Compatibilityevaluationforthreeprimer-amplificationandsevenanalyticalapproaches,usingamockmicrobialcommunitysample.(A)Beta-diversityordination(Bray–Curtisdissimilarity)ofthehumanfecalmicrobiome(EP1–EP10;coloredsymbols)andmockcommunity(M;blacksymbols)sequencingsamples.Notation:3,V3V4ampliconlibrary;O,successfulassemblyofV4Oampliconlibrary;x,failedassemblyofV4Oampliconlibrary;N,V4Nampliconlibrary;p,V4OSE,single-endanalysisusingV4Oampliconlibrary;q,V4NSE,single-endanalysisusingV4Nampliconlibrary;o,tV4O,trimmedV4OampliconfromV3V4library;n,tV4N,trimmedV4NampliconfromV3V4library.Themockcommunitysampleanalyzedbysevenanalyticalapproachesclusteredtogetherandawayfromdatafor10individualhumanfecalmicrobiomes.(B)Beta-diversityordinationofthesevenanalyticalapproachesusedfortheanalysisofmockcommunitydata.(C)Therelativeabundanceofbacteriainthemockcommunity,comparingthetheoreticalcompositionwiththatobtainedbysevenanalyticalapproaches.First,wecomparedthedataformockcommunitysamples(simplesamplecomposition)withhumanfecalsamples(complexsamplecomposition)usingthesevenanalyticapproaches(Fig.3A).Thebeta-diversity(measuredbyBray–Curtisdissimilarity)ordinationplotrevealedhighhomogeneityinmockcommunityquantificationofthesevenapproaches.TheBray–Curtisdissimilarityisaquantitativemeasurement.Accordingly,azoomed-inviewofthebeta-diversityordinationforthemockcommunityrevealedthatthePCRconditions(primersused)reflectedthecommunityquantitativecomposition(Fig.3B).Specifically,thetrimmedV4datapointswereclosetotheV3V4datapoint;thePEandSEV4Odatapointswereclusteredtogether;andthePEandSEV4Ndatapointswereproximalonthefirstplotaxis.Therelativetaxonomicabundancesinthetheoreticalsample,asdeterminedbythesevenanalyticalapproaches,werenotsignificantlydifferent(G-test;meanofG=1.31,P=0.99;Fig.3C).Theratesofmisclassification(unassignedorassignedtoyeasts)werelessthan1%inallanalyses.Thedifferencesinthedeterminedrelativeabundanceandthetheoreticalcompositionrangedfrom8.14%(V4O)to15.22%(V4N).SamplecompositiondeterminedbytheV3V4approach(anditsderivedV4approaches)differedbyapproximately10%fromthetheoreticalcomposition.Thisindicatesthatthedifferent16Sregionsandanalyticapproachesforprofilingasimplecompositionmicrobiota,suchasasynthesizedmockcommunity,areabletopresentsimilarquantitativeandqualitativeresults.Determinationofsampleheterogeneityandvariationoftaxacompositionafter16SamplificationwithdifferentprimersWenextanalyzedhumanfecalsamplestoevaluatethecomplementarityofdifferent16Samplificationanalyticalapproaches.Thesevenanalyticalapproachesusedwerethesameasthoseforthemockcommunityanalysis.Weexaminedthebetadiversity,asdeterminedbythesevenapproaches(Fig.4).Thebeta-diversityprofilesforfecalmicrobiotadatapointsforthesevenapproachesreflectedthe10individualsamplesources(Fig.4A).Wetestedthesampleandanalyticalapproachheterogeneitybytheanalysisofsimilarities(ANOSIM).Thebetween-samplevariancewasgreaterthanwithin-samplevariance(R=0.987,P=0.001;Fig.4AandFig.4B);ontheotherhand,thebetween-analyticalapproachvariancewasnotsignificantlygreaterthanthewithin-analyticalapproachvariance(R=0.016,P=0.22;Fig.4CandFig.4D).Inotherwords,thedeterminedsamplevariationwasrelativelyconstant,regardlessoftheprimersoranalyticalapproachesused.DownloadfigureOpeninnewtabFIG4Compatibilityevaluationofthreeprimer-amplificationandsevenanalyticalapproachesbytestingtheheterogeneityof10humanfecalmicrobiomes.(A)Beta-diversityordination(Bray–Curtisdissimilarity)of10humanfecalmicrobiomesamples(colored-borderpolygons,EP1–EP10).(B)Nestedanalysisofsimilarity(ANOSIM)ofindividualtestingapproaches,nestedbytheanalyticalapproach;R=0.987,P=0.001.(C)Beta-diversityordinationof10humanfecalmicrobiomesamples(radiallines;theanalyticalapproachesarelabeledinthecenter).(D)NestedANOSIMoftheanalyticalapproachesnestedbyindividuals;R=0.016,P=0.22.AbbreviationsareasinFig.3.Wenextanalyzedthreealpha-diversityindices(namely,richness,Shannonentropy,andSimpsonindexofdiversity),asdeterminedbyusingthesevenanalyticalapproaches(Fig.S1).Thethreeindiceswerenotstatisticallysignificantlydifferentwhenthedifferentapproacheswereused(P>0.05).However,thedifferencesintherichnessindexweremarginallysignificant(P=0.08)forthesevenapproaches;thistrendwasattributedtoreducedtaxonnumbersinsamplesanalyzedusingthetrimmedV4OandtrimmedV4Napproaches(Fig.S1A).Consistencyoftaxonomicabundancesdeterminedbydifferent16SampliconanalyticalapproachesPCRartifacts(over-amplifiedampliconsandchimeras)interferewithtaxonomicquantificationandtaxonomicstructureprofilingofmicrobiomesamples.TheartifactsaresequencesthatareamplifiedinabiasedmannerduringPCR.Nouniversalprimerexistsforafullyunbiasedamplification.Inthecurrentstudy,weusedtheDADA2denoisingpipelinetoreducetheconfoundingeffectofPCRartifacts.Weprofiledthehigher-leveltaxonomycompositions(therelativephylumabundance)byusingstackedbarplots(Fig.5A).Theproportionofeachphylumwasdifferentfordifferentanalyticalapproachused,butthedifferencewasnotstatisticallysignificant(G-test;meanofG=0.81,P=1).However,theV4N-PEanalyticalapproachunder-detectedthephylumVerrucomicrobia(relativeabundance0.02%vs.1%ofotherapproaches).DownloadfigureOpeninnewtabFIG5Phylumcompositionsandtaxonomicaccuracyofthehumanfecalmicrobiomesanalyzedbysevenanalyticalapproaches.(A)Phylumcompositionsofthehumanfecalmicrobiome(10individualsaverageperstackedbar)analyzedbysevenanalyticalapproaches.(B)PhylumtospeciestaxonomicaccuraciesbasedontheV3V4taxonomyassignment.(C)PhylumtospeciestaxonomicaccuraciesbasedontheV4Otaxonomyassignment.AbbreviationsareasinFig.3.Becausenotheoreticalreferenceexistsforthehumanfecalmicrobiotacomposition,weusedtheV3V4approach(thelongestsequenceregion;Fig.5B)andtheV4Oapproach(themostaccuratemethodbasedoninsilicosimulationandmockcommunityexperiments;Fig.5C)asbenchmarkstoevaluatetheclassificationaccuracy.ThetaxonomydataobtainedbyV3V4-derivedanalyses(trimmedV4OandtrimmedV4N)weretheclosesttothoseoftheV3V4approachabovethegenuslevel(Fig.5B).ComparedwiththeV3V4data,theaccuracyofV4OandV4N(PEandSE)methodsabovethegenuslevelwas78.9%to91.3%.Ontheotherhand,comparedwiththeV4Odata,thetaxonomydeterminedbytheV4OSEmethodwastheclosesttothatoftheV4O(PE)method(Fig.5C).Theothermethodsreachedan83.1%to93.0%averagerelativeaccuracyabovethegenuslevel.Therankabundanceanalysisatthefamily(Fig.S2A)andgenuslevels(Fig.S2B)revealedthatthetaxonabundancewasconsistentwiththeV4O-basedclassificationaccuracy.Wethenperformedpairwiseanalysisoftaxonomicabundancecorrelationsbetweenthesevenanalyticalapproaches(Fig.6).Weplottedtheabundanceofeachtaxonomy-assignedASVusingpairwise-correlationscatterplots(Fig.6,lowerleftpanels)andnotedtheabundancecorrelationcoefficients(Fig.6,upperrightpanels).ThecorrelationcoefficientsbetweentheV3V4dataandothermethodsrangedfrom0.54to0.69.ThecorrelationcoefficientsbetweentheV4methoddata,regardlessofwhetherthesewerePE,SE,ortrimmedmethods,andtheV4OandV4Ndatawerehigh(0.92to0.94).Inaddition,fortheV4PEdata,weobservedthebestcorrelationbetweenthecorrespondingtrimmedV4methods(0.99forV4OandtV4O;and0.98forV4NandtV4N)(summarizedinTable3).AhighconsistencyamongV4approachespresentedbyintegratingbothtaxonomicassignmentandquantification,andtheV3V4ampliconswouldbecompatiblewithV4bytrimminguptothesamehypervariableregion.DownloadfigureOpeninnewtabFIG6Taxonomicabundancecorrelationsofsevenanalyticalapproaches.Lowerleftpanels,pairwisetaxonomicabundancecorrelations.Eachpointisanampliconsequencevariantsharedbyanytwo-analysissets.Upperrightpanels,Pearson’scorrelationcoefficientsforpairwisetaxonomicabundancecorrelations;alltestsweresignificant.AbbreviationsareasinFig.3.DISCUSSIONTheincreasingpopularityandadvancesinsequencingtechnologyprompthumanandclinicalmicrobiomestudies.However,thetypesofsequencingtechnologyandreagentsusedlimitmeaningfulcomparisonsoftheobtaineddata.Here,wecomparedthemostpopularmicrobiomesequencingapproachesthatrelyontheanalysisof16Samplicons(i.e.,amplifyingdifferenthypervariableregionsandcouplingwithDADA2sequencevariants’denoisingprocess).Weevaluatedtheaccuracyandconsistencyof16Samplicons,whicharecapturedbyV3V4andV4regions.OurfindingsshowedthatV4O(515F-806R)primeryieldedthemostconsistent,complementary,andaccuratetaxonomicprofile.Additionally,weimplementedanintegratedapproachfortheV3V4andV4ampliconsfromdifferentdatasetsbytrimmingV3V4sequencesuptotheV4region.TheNGStechnologyandHMP(HMP1andiHMP)generatemassivesequencingdata,promotingclinicalmicrobiomeassociationstudies(33).Thesestudiesrevealnumerouspreviouslyunknownhost–microbeinteractionsthatunderlievarioushealthissues,especiallynon-communicablediseases,withnoveletiologylinkages(3-7).Forapopulation-basedcohortstudyintheclinicalmicrobiomeresearch,datathroughputandquality,thecost–benefitaspect,andvalidityoftheanalyticalpipelineshouldallbeconsidered.The16SV3V4andV4hypervariableregionsarewidelyselectedforhumanmicrobiotaprofiling,butthefragmentsamplifiedfromdifferentregionsshouldbecoupledwithsuitablesequencingconditions.Accordingly,wehereevaluatedthecompatibilityofdifferentampliconlibrariesbyinsilicoandempiricalapproaches,coupledwithadenoisingalgorithmtodetectASVs.Theanalysisdemonstratedthatthecurrentwidelyusedampliconprimerscaptureover80%ofthetaxonomicinformation.Theampliconsequencelengthwasacriticalfactorfortaxonomicprofilingaslongersequencescontainmoregeneticinformationthanshortersequences(11).Theprimerswereanothercriticalfactor,astheampliconsaremorerepresentativeofthesamplecomplexitywhengeneratedbymoreuniversal(conserved)primers.Forexample,whiletheV3V4andV4Oapproachessharenearlyidenticalreverseprimers,theV3V4librarywascharacterizedbyahigherASVrichnessindex(ASVnumbers;Fig.S1A)butcontainedlesstaxonomicinformationthantheV4Olibrary(thearchaealphylawerealmostabsentintheV3V4library;Fig.2AandTable3).Thetaxonomicassignmentaccuracyiscrucialforthesubsequentstudydesign,suchasstrainisolation,identification,andclinicalorcommercialapplications(12).TheinsilicoPCRanalysisperformedhereindemonstratedthattheV4OtaxonomyassignmentismoreaccuratethantheV3V4-basedassignments.However,eventhoughtheV4N-capturedsequencesoverlappedwithmostregionscapturedbytheV4Omethod,thetaxonomicaccuracy(atmost70%accuracyatphylumlevel)wasmuchlowerthanthatofeitherV3V4orV4Omethod.Inpractice,thetV4Oapproach(V4OtrimmedfromV3V4)wascompatibleforintegratingV3V4-generateddatawithV4O-generateddata(Fig.6)whentestingthehumangutmicrobiomesamples,whichcontainfewerarchaeathanenvironmentalsamples.Sinceshort-readampliconsequencingonlyextractspartialinformationforthe1.5-kb16SrRNAgene,thetaxonomicidentificationisatmostlimitedatapproximately80%accuracyatthegenuslevelbyV3V4primeramplificationandatthespecieslevelbyV4Oprimeramplification.TheV4ampliconsequencingmeetseconomicbenefitsformicrobiomestudies.Ampliconsequencinginmanystudies,notonlyenvironmentalecologicalstudiesbutalsohuman-associatedstudies,focusesontheV4regionandreliesontheEMPprotocol.However,thelibraryconstructionmethodsarerestrictedbythemaximumreadlengthofasequencer(30).Inotherwords,theamplifiedV4ampliconfragmentsize(PCRproductwithouttheadapter-linkingsequences)isexpectedtobeapproximately291bpifthelibraryisconstructedbyfollowingtheNexteraXTtwo-stepdualindexPCRprotocol.Thetwo-stepdualindexPCRmethodwasofficiallydesignedfor16SV3V4libraryconstruction,coupledwithMiSeq600-cycle(300PE)sequencingbyIllumina(15).Althoughthetwo-stepdualindexPCRisnotexclusivetoMiSeqandV3V4amplicons(34),itssequencingoutputs(i.e.,thereadlengthandPEoverlap)shouldbepreciselycalculatedbasedonthequalityscoredistributionsfromthe5’-endtothe3’-endoftheread(35).Forexample,thePEreadsoverlapbylessthan10bpwhenaV4ampliconlibraryisconstructedbyusingNexteraXTkitandsequencedas150PE.MostV4150PEreadsdonotpassthequalityfilteringandcannotbeusedtoassemblebothends.Inaddition,theDADA2algorithmonlyacceptsPEreadswithamorethan20-bpoverlapasdefault(24).Caporasoetal.(36)developedcustomizedsequencingprimersbasedonHiSeq150PE.Theensuingcustomizedsequencingproceduresentailedmodificationofthelibraryconstructionmethods(forwardprimersweredirectlylinkedtobarcodesequencesandanadapter),thesequencingprogram(usingtheTruSeqworkflowandignoringanerrormessagefromthesequencersystem),andincreasingtherunbyadditional14sequencingcycles.Consequently,themethodnotonlysolvedtheunassembledreadproblem(yielding253-bpampliconfragments)butalsoloweredthesequencingcostpersample(Table1).However,eventhoughinsomestudies,theearlystagesofsequencingdonotfollowtheEMPorCaporasoprotocols,theunassembledreadscanbeanalyzedbyusingasingle-endpipeline.Weheredemonstratedthatthesingle-endpipelinewascomparablewiththePEpipeline,withmorethan90%confidenceinthetaxonomyassignmentbutaslightlyinferiorquantification(Pearson’scorrelationcoefficient0.75)(Table3).TheASVdenoisingmethods(23)(e.g.,DADA2(24),Deblur(25),andUNOISE3(27))arereplacingthetraditionaloperationaltaxonomicunitclusteringmethodsforchoosingrepresentativeampliconsequences.Thesedenoisingmethodsdetectrealbiologicalsequencefeaturesmissedbyclustering,anddenoisedfeaturesarespecificandreproducible(23).Therefore,wesuggesttrimmingtheV3V4andV4sequencestothesameregiontoacquirethesamerepresentativeASVswhencombiningdifferentlibraries.AlthoughtheV3V4approachunder-detectedthearchaea,theanalysisoftrimmedV4ampliconsrecoveredthequalitativeandquantitativeaspectsofthehumangutmicrobiomewithover90%confidence(Table3andFig.6).Inconclusion,inthecurrentstudy,weevaluatedthecompatibilityof16SV3V4andV4ampliconstypicallyanalyzedinclinicalmicrobiomeprofilingstudiesbyusingasequencevariant-denoisingpipeline.Ourfindingssuggestthat:(1)theanalysisofthePEV4Oamplicon(amplifiedusingprimers515F–806R)resultsinthemostaccuratetaxonomicassignment;(2)theV3V4ampliconanalysisiscompatiblewiththeV4ampliconanalysisaftertrimmingtothesameregion;and(3)whilemid-highthroughputsequencersreducethecostofsequencingpersample,onlyacustomizedV4libraryissuitableforstitchingPEreadsforsubsequentanalyses(36).Thefindingsareempiricalandanalyticalsuggestionsforcost-effectivepopulation-basedormeta-analysisclinicalmicrobiomestudies.MATERIALSANDMETHODSEthicsstatementandsamplecollectionThestudiesinvolvinghumanfecalsamplecollectionandinformedconsentfromhumanparticipantswereapprovedbytheInstitutionalReviewBoardofNationalTaiwanUniversityHospital,Taipei,Taiwan(201606045RINB).Fecalsamplesfrom10healthyvolunteerswerecollectedduringFebruary2017attheNationalTaiwanUniversity,asdescribedbyWuetal.(37).PrimerselectionandalternativeV4primerdesignTwosetsof16Sampliconprimers,whichtargetedtheV3V4(primers341F–805R)andV4(primers515F–806R;V4O)regions,wereselectedfromtheIllumina-recommend(38)andEMPprotocols(18),respectively.AnalternativepairofV4region-specificprimers(V4N)consistedofmodifiedprimersofEMPV4andGhyselincketal.(39):519F,5’-CAGCMGCCGCGGTAAT-3’,and798R,5’-GGGTWTCTAATCCKGTT-3’.TheexpectedPCRproductlengthwas279bp.TheV4NprimersweresynthesizedwithoverhangadaptersforindexattachmentandIlluminasequencingadapters(15).ThecoveragerateofeachprimerpairwasevaluatedbyusinginsilicoPCRsimulation(seebelow).InsilicoPCRsimulationThereadandtaxonomycoverageratesoftheSILVA16Sgenedatabase(NR13299%)(31,32)wereevaluatedbyinsilicoPCR.ThesimulationwasconductedbyextractingtheexpectedPCRfragmentsgeneratedbyamplificationusingtheV3V4,V4O,andV4Nprimers.TheinsilicoPCRpipelinewasset-upusingaUNIXshellscript,asfollows:(1)createalistofdegenerateprimerpairs,andlinktheforwardprimertothereverseprimerwith“.*”fromthe5’-endtothe3’-end(e.g.,CCTACGGGAGGCAGCAG.*GGATTAGATACCCCAGTAGTC);(2)countthefragmentsextractedfromthedatabasefastafilebyusingUNIXcommand“grep”withtheparameter-c;(3)obtainthetargetedsequencefragmentsusingtheparameter-o;(4)extractthesequenceIDusingtheparameter-B1;(5)countthelengthofinsilicoPCRproductsusingtheawk‘{printlength}’UNIXcommand.SequencinglibrarypreparationformockcommunityandhumanfecalmicrobiomesDNAextractionGenomicDNAfromamockcommunitystandard(ZymoBIOMICSMicrobialCommunityStandard,catalogno.D6300,ZYMORESEARCH,CA,USA)andfromstoolsamplesfrom10volunteerswasextractedusingQIAamp®PowerFecal®DNAKit(QIAGEN,catalogno.12830–50;Hilden,Germany).ThegenomicDNAwasstoredatm–20°CuntilPCRamplificationandampliconsequencing.AmplificationandNGSsequencingTwo-stepPCRwasperformed,followingtheIlluminaprotocolfor16Smetagenomicsequencinglibrarypreparation.PCRwasfirstperformedtocapturethe16SV3V4(primers341F–805R)andV4hypervariableregions(primers515F–806Rand519F–798R).ThreelibrarieswerethenconstructedbyindexPCRusingtheNexteraXTduel-IndexPCRprimers(15).ThepooledlibrarieswerePE-sequencedinthesamerunusingtheIlluminaMiSeqreagentkitversion3(SanDiego,CA,USA)for600cyclesattheMedicalMicrobiotaCenteroftheFirstCoreLaboratory,NationalTaiwanUniversityCollegeofMedicine.BioinformaticanalysisformicrobialtaxonomicprofilingSequencedenoisingusingDADA2andtheQIIME2pipelineThesequenceswereprocessedbyusingQIIME2pipeline(version2019.10)(9).Theprimersequencesweretrimmedfromtherawreadsinthethreelibrariesbyusingthecutadaptplugin.Thetrimmedsingle-endorPEsequencesweresubsequentlydenoisedusingtheDADA2plugininQIIME2.ToobtainqualifiedASVs,thereadsweretruncatedfromthe3’-endbasedonthequalityscoredistributiontothefollowingreadlength:(1)V3V4-forward,270bp,andV3V4-reverse,210bp;(2)V4O-forward,131bp,andV4O-reverse,130bp;and(3)V4N-forward,134bp,andV4N-reverse,133bp.Inaddition,theV3V4readsweretrimmedtotheV4OandV4Nread-lengthforfurthercomparisons.High-confidenceASVswerethenobtainedbydenoising,withqualityfilteringandchimeraremoval.ThetaxonomywasassignedusinganaÏveBayesclassifiertrainedontheSILVA13299%full-length16SrRNAgenesequencedatabase(31,32).MicrobialbiodiversityandstatisticalanalysesAllstatisticalanalyseswereconductedwithRversion3.6.1(40).MicrobialcommunityanalyseswereperformedusingtheveganRpackage(41).Alphadiversityindices,includingtheShannonindexandSimpsonindex,weredeterminedbyusingthe“diversity”function;theRichnessindexwascalculatedbyusingthe“specnumber”function.ThebetadiversitywasdeterminedbasedontheBray–Curtisdissimilarityandvisualizedbyprincipalcoordinatesanalysis.ANOSIMwasusedtotesttheheterogeneityamongindividuals,controllingforthedifferentprimers.AllunivariateanalyseswereconductedbytheKruskal–Wallistestwithα=0.05cut-offforsignificanceandDunn’stestforpost-hoccomparisons.Multiple-testingP-valueswereadjustedbasedonthefalsediscoveryratebyusingthe“p.adjust”functioninR.Thelikelihood-ratiotest(G-testwithα=0.05cut-offforsignificance)fortheabundanceprofileswasperformedbyusingtheRVAideMemoireRpackage.DataavailabilitySequencesgeneratedinthecourseofthecurrentstudyhavebeendepositedintheSequenceReadArchive(SRA)databaseundertheaccessionnumberPRJNA643648.SUPPLEMENTARYFIGURELEGENDSSupplementaryFiguresDownloadfigureOpeninnewtabFIGS1Alpha-diversityindicesobtainedwithsevenanalyticalapproachesbytestingtheheterogeneityof10humanfecalmicrobiomes.(A)Richness(ampliconsequencevariantnumber).(B)Shannonentropy.(C)Simpsonindexofdiversity.AllindicesweretestedbyKruskal–Wallisranksumtestwithalpha=0.05.Thesevenanalyticalapproacheswereasfollows:paired-endV3V4amplicons(V3V4),paired-endV4Oamplicons(V4O),paired-endV4Namplicons(V4N),single-endV4Oamplicons(V4OSE),single-endV4Namplicons(V4NSE),andV4ampliconstrimmedfromV3V4(trimmedV4O,tV4O;andtrimmedV4N,tV4N).DownloadfigureOpeninnewtabFIGS2Taxonomicrankabundancesobtainedwithsevenanalyticalapproachesbytestingtheheterogeneityof10humanfecalmicrobiomes.(A)Familyrankabundance.Barfillcolorcorrespondstothephylum.(B)Genusrankabundance.AbbreviationsareasinFig.S1.SupplementaryTableViewthistable:ViewinlineViewpopupDownloadpowerpointTABLES1In-silicoarchaealsequencecaptureratesoftheV3V4andV4primersACKNOWLEDGMENTSWethanktheresearchparticipantsandresearchassistantsfromtheInstituteofFoodScienceandTechnology,NationalTaiwanUniversity,Taipei,Taiwan(Guan-LingOu)andNationalTaiwanUniversityCollegeofMedicine,Taipei,Taiwan(Yu-TangYangandFang-WeiKuo).WewouldalsoliketoacknowledgethesequencingserviceprovidedbytheMedicalMicrobiotaCenteroftheFirstCoreLaboratory,NationalTaiwanUniversityCollegeofMedicine,thecomputationalresourcesupportbyProf.AlexHon-TsenYuattheDepartmentofLifeScience,NationalTaiwanUniversity,andtechnicalconsultingbyAn-ChiChengattheUniversityofFlorida,Gainesville,FL,USA.Wedeclarethatwehavenocompetinginterests.P.Y.L.andW.K.W.conceivedandplannedtheproject.W.K.W.,C.C.C.,andS.P.wereinvolvedinsamplecollection,processing,andstorage,andsupervisedtheexperiments.P.Y.L.conductedallbioinformaticandstatisticalanalyses.P.Y.L.andW.K.W.wereinvolvedindatainterpretationandmanuscriptplanning.P.Y.L.draftedthemanuscript.L.Y.S.andM.S.W.supervisedthestudy.Allauthorsapprovedsubmissionofthefinalversion.REFERENCES1.↵LeyRE,TurnbaughPJ,KleinS,GordonJI.2006.Microbialecology:humangutmicrobesassociatedwithobesity.Nature444:1022–3.OpenUrlCrossRefPubMedWebofScience2.↵TurnbaughPJ,LeyRE,MahowaldMA,MagriniV,MardisER,GordonJI.2006.Anobesity-associatedgutmicrobiomewithincreasedcapacityforenergyharvest.Nature444:1027–31.OpenUrlCrossRefPubMedWebofScience3.↵ZhuW,GregoryJC,OrgE,BuffaJA,GuptaN,WangZ,LiL,FuX,WuY,MehrabianM,SartorRB,McIntyreTM,SilversteinRL,TangWHW,DiDonatoJA,BrownJM,LusisAJ,HazenSL.2016.GutMicrobialMetaboliteTMAOEnhancesPlateletHyperreactivityandThrombosisRisk.Cell165:111–124.OpenUrlCrossRefPubMed4.↵SampsonTR,DebeliusJW,ThronT,JanssenS,ShastriGG,IlhanZE,ChallisC,SchretterCE,RochaS,GradinaruV,ChesseletMF,KeshavarzianA,ShannonKM,Krajmalnik-BrownR,Wittung-StafshedeP,KnightR,MazmanianSK.2016.GutMicrobiotaRegulateMotorDeficitsandNeuroinflammationinaModelofParkinson’sDisease.Cell167:1469–1480e12.OpenUrlCrossRefPubMed5.↵BrennanCA,GarrettWS.2016.GutMicrobiota,Inflammation,andColorectalCancer.AnnuRevMicrobiol70:395–411.OpenUrlCrossRefPubMed6.↵Lloyd-PriceJ,ArzeC,AnanthakrishnanAN,SchirmerM,Avila-PachecoJ,PoonTW,AndrewsE,AjamiNJ,BonhamKS,BrislawnCJ,CaseroD,CourtneyH,GonzalezA,GraeberTG,HallAB,LakeK,LandersCJ,MallickH,PlichtaDR,PrasadM,RahnavardG,SaukJ,ShunginD,Vazquez-BaezaY,WhiteRA,3rd.,InvestigatorsI,BraunJ,DensonLA,JanssonJK,KnightR,KugathasanS,McGovernDPB,PetrosinoJF,StappenbeckTS,WinterHS,ClishCB,FranzosaEA,VlamakisH,XavierRJ,HuttenhowerC.2019.Multi-omicsofthegutmicrobialecosystemininflammatoryboweldiseases.Nature569:655–662.OpenUrlCrossRefPubMed7.↵HoylesL,Fernandez-RealJM,FedericiM,SerinoM,AbbottJ,CharpentierJ,HeymesC,LuqueJL,AnthonyE,BartonRH,ChillouxJ,MyridakisA,Martinez-GiliL,Moreno-NavarreteJM,BenhamedF,AzalbertV,Blasco-BaqueV,PuigJ,XifraG,RicartW,TomlinsonC,WoodbridgeM,CardelliniM,DavatoF,CardoliniI,PorzioO,GentileschiP,LopezF,FoufelleF,ButcherSA,HolmesE,NicholsonJK,PosticC,BurcelinR,DumasME.2018.Molecularphenomicsandmetagenomicsofhepaticsteatosisinnon-diabeticobesewomen.NatMed24:1070–1080.OpenUrlCrossRefPubMed8.↵McFall-NgaiM,HadfieldMG,BoschTC,CareyHV,Domazet-LosoT,DouglasAE,DubilierN,EberlG,FukamiT,GilbertSF,HentschelU,KingN,KjellebergS,KnollAH,KremerN,MazmanianSK,MetcalfJL,NealsonK,PierceNE,RawlsJF,ReidA,RubyEG,RumphoM,SandersJG,TautzD,WernegreenJJ.2013.Animalsinabacterialworld,anewimperativeforthelifesciences.ProcNatlAcadSciUSA110:3229–36.OpenUrlAbstract/FREEFullText9.↵BolyenE,RideoutJR,DillonMR,BokulichNA,AbnetCC,Al-GhalithGA,AlexanderH,AlmEJ,ArumugamM,AsnicarF,BaiY,BisanzJE,BittingerK,BrejnrodA,BrislawnCJ,BrownCT,CallahanBJ,Caraballo-RodriguezAM,ChaseJ,CopeEK,DaSilvaR,DienerC,DorresteinPC,DouglasGM,DurallDM,DuvalletC,EdwardsonCF,ErnstM,EstakiM,FouquierJ,GauglitzJM,GibbonsSM,GibsonDL,GonzalezA,GorlickK,GuoJ,HillmannB,HolmesS,HolsteH,HuttenhowerC,HuttleyGA,JanssenS,JarmuschAK,JiangL,KaehlerBD,KangKB,KeefeCR,KeimP,KelleyST,KnightsD,etal.2019.Reproducible,interactive,scalableandextensiblemicrobiomedatascienceusingQIIME2.NatBiotechnol37:852–857.OpenUrlCrossRefPubMed10.↵ComeauAM,DouglasGM,LangilleMG.2017.MicrobiomeHelper:aCustomandStreamlinedWorkflowforMicrobiomeResearch.mSystems2.11.↵JohnsonJS,SpakowiczDJ,HongBY,PetersenLM,DemkowiczP,ChenL,LeopoldSR,HansonBM,AgrestaHO,GersteinM,SodergrenE,WeinstockGM.2019.Evaluationof16SrRNAgenesequencingforspeciesandstrain-levelmicrobiomeanalysis.NatCommun10:5029.OpenUrlCrossRef12.↵BokulichNA,KaehlerBD,RideoutJR,DillonM,BolyenE,KnightR,HuttleyGA,GregoryCaporasoJ.2018.Optimizingtaxonomicclassificationofmarker-geneampliconsequenceswithQIIME2’sq2-feature-classifierplugin.Microbiome6:90.OpenUrlCrossRefPubMed13.↵WuWK,ChenCC,LiuPY,PanyodS,LiaoBY,ChenPC,KaoHL,KuoHC,KuoCH,ChiuTHT,ChenRA,ChuangHL,HuangYT,ZouHB,HsuCC,ChangTY,LinCL,HoCT,YuHT,SheenLY,WuMS.2019.IdentificationofTMAO-producerphenotypeandhost-diet-gutdysbiosisbycarnitinechallengetestinhumanandgerm-freemice.Gut68:1439–1449.OpenUrlAbstract/FREEFullText14.↵GohlDM,VangayP,GarbeJ,MacLeanA,HaugeA,BeckerA,GouldTJ,ClaytonJB,JohnsonTJ,HunterR,KnightsD,BeckmanKB.2016.Systematicimprovementofampliconmarkergenemethodsforincreasedaccuracyinmicrobiomestudies.NatBiotechnol34:942–9.OpenUrlCrossRefPubMed15.↵Illumina.2013.Illumina16Smetagenomicsequencinglibrarypreparation(IlluminaTechnicalNote15044223Rev.A).http://support.illumina.com/content/dam/illumina-support/documents/documentation/chemistry_documentation/16s/16s-metagenomic-library-prep-guide-15044223-b.pdf.Accessed2018/10.16.↵HumanMicrobiomeProjectC.2012.Structure,functionanddiversityofthehealthyhumanmicrobiome.Nature486:207–14.OpenUrlCrossRefPubMedWebofScience17.↵SinclairL,OsmanOA,BertilssonS,EilerA.2015.Microbialcommunitycompositionanddiversityvia16SrRNAgeneamplicons:evaluatingtheilluminaplatform.PLoSOne10:e0116955.OpenUrlCrossRefPubMed18.↵ThompsonLR,SandersJG,McDonaldD,AmirA,LadauJ,LoceyKJ,PrillRJ,TripathiA,GibbonsSM,AckermannG,Navas-MolinaJA,JanssenS,KopylovaE,Vazquez-BaezaY,GonzalezA,MortonJT,MirarabS,ZechXuZ,JiangL,HaroonMF,KanbarJ,ZhuQ,JinSongS,KosciolekT,BokulichNA,LeflerJ,BrislawnCJ,HumphreyG,OwensSM,Hampton-MarcellJ,Berg-LyonsD,McKenzieV,FiererN,FuhrmanJA,ClausetA,StevensRL,ShadeA,PollardKS,GoodwinKD,JanssonJK,GilbertJA,KnightR,EarthMicrobiomeProjectC.2017.AcommunalcataloguerevealsEarth’smultiscalemicrobialdiversity.Nature551:457–463.OpenUrlCrossRefPubMed19.↵GilbertJA,JanssonJK,KnightR.2018.EarthMicrobiomeProjectandGlobalSystemsBiology.mSystems3.20.↵CaporasoJG,LauberCL,CostelloEK,Berg-LyonsD,GonzalezA,StombaughJ,KnightsD,GajerP,RavelJ,FiererN,GordonJI,KnightR.2011.Movingpicturesofthehumanmicrobiome.GenomeBiol12:R50.OpenUrlCrossRefPubMed21.↵McDonaldD,HydeE,DebeliusJW,MortonJT,GonzalezA,AckermannG,AksenovAA,BehsazB,BrennanC,ChenY,DeRightGoldasichL,DorresteinPC,DunnRR,FahimipourAK,GaffneyJ,GilbertJA,GogulG,GreenJL,HugenholtzP,HumphreyG,HuttenhowerC,JacksonMA,JanssenS,JesteDV,JiangL,KelleyST,KnightsD,KosciolekT,LadauJ,LeachJ,MarotzC,MeleshkoD,MelnikAV,MetcalfJL,MohimaniH,MontassierE,Navas-MolinaJ,NguyenTT,PeddadaS,PevznerP,PollardKS,RahnavardG,Robbins-PiankaA,SangwanN,ShorensteinJ,SmarrL,SongSJ,SpectorT,SwaffordAD,ThackrayVG,etal.2018.AmericanGut:anOpenPlatformforCitizenScienceMicrobiomeResearch.mSystems3.22.↵CaporasoJG,LauberCL,WaltersWA,Berg-LyonsD,LozuponeCA,TurnbaughPJ,FiererN,KnightR.2011.Globalpatternsof16SrRNAdiversityatadepthofmillionsofsequencespersample.ProcNatlAcadSciUSA108Suppl1:4516–22.OpenUrlCrossRef23.↵NearingJT,DouglasGM,ComeauAM,LangilleMGI.2018.DenoisingtheDenoisers:anindependentevaluationofmicrobiomesequenceerror-correctionapproaches.PeerJ6:e5364.OpenUrlCrossRef24.↵CallahanBJ,McMurdiePJ,RosenMJ,HanAW,JohnsonAJ,HolmesSP.2016.DADA2:High-resolutionsampleinferencefromIlluminaamplicondata.NatMethods13:581–3.OpenUrlCrossRefPubMed25.↵AmirA,McDonaldD,Navas-MolinaJA,KopylovaE,MortonJT,ZechXuZ,KightleyEP,ThompsonLR,HydeER,GonzalezA,KnightR.2017.DeblurRapidlyResolvesSingle-NucleotideCommunitySequencePatterns.mSystems2.26.↵EdgarRC,FlyvbjergH.2015.Errorfiltering,pairassemblyanderrorcorrectionfornext-generationsequencingreads.Bioinformatics31:3476–82.OpenUrlCrossRefPubMed27.↵EdgarRC.2016.UNOISE2:improvederror-correctionforIllumina16SandITSampliconsequencing.bioRxivdoi:10.1101/081257:081257.OpenUrlCrossRef28.↵BolyenE,RideoutJR,DillonMR,BokulichNA,AbnetC,Al-GhalithGA,AlexanderH,AlmEJ,ArumugamM,AsnicarF,BaiY,BisanzJE,BittingerK,BrejnrodA,BrislawnCJ,BrownCT,CallahanBJ,Caraballo-RodríguezAM,ChaseJ,CopeE,DaSilvaR,DorresteinPC,DouglasGM,DurallDM,DuvalletC,EdwardsonCF,ErnstM,EstakiM,FouquierJ,GauglitzJM,GibsonDL,GonzalezA,GorlickK,GuoJ,HillmannB,HolmesS,HolsteH,HuttenhowerC,HuttleyG,JanssenS,JarmuschAK,JiangL,KaehlerB,KangKB,KeefeCR,KeimP,KelleyST,KnightsD,KoesterI,KosciolekT,etal.2018.QIIME2:Reproducible,interactive,scalable,andextensiblemicrobiomedatascience.PeerJPreprints6.29.↵WetterstrandKA.DNASequencingCosts:DatafromtheNHGRIGenomeSequencingProgram(GSP)www.genome.gov/sequencingcostsdata.Accessed2020/07/06.30.↵HolmJB,HumphrysMS,RobinsonCK,SettlesML,OttS,FuL,YangH,GajerP,HeX,McCombE,GravittPE,GhanemKG,BrotmanRM,RavelJ.2019.Ultrahigh-ThroughputMultiplexingandSequencingof>500-Base-PairAmpliconRegionsontheIlluminaHiSeq2500Platform.mSystems4.31.↵QuastC,PruesseE,YilmazP,GerkenJ,SchweerT,YarzaP,PepliesJ,GlocknerFO.2013.TheSILVAribosomalRNAgenedatabaseproject:improveddataprocessingandweb-basedtools.NucleicAcidsRes41:D590–6.OpenUrlCrossRefPubMedWebofScience32.↵YilmazP,ParfreyLW,YarzaP,GerkenJ,PruesseE,QuastC,SchweerT,PepliesJ,LudwigW,GlocknerFO.2014.TheSILVAand“All-speciesLivingTreeProject(LTP)”taxonomicframeworks.NucleicAcidsResearch42:D643–D648.OpenUrlCrossRefPubMedWebofScience33.↵IntegrativeHMPRNC.2014.TheIntegrativeHumanMicrobiomeProject:dynamicanalysisofmicrobiome-hostomicsprofilesduringperiodsofhumanhealthanddisease.CellHostMicrobe16:276–89.OpenUrlCrossRefPubMed34.↵GlennTC,PiersonTW,Bayona-VasquezNJ,KieranTJ,HoffbergSL,ThomasIvJC,LefeverDE,FingerJW,GaoB,BianX,LouhaS,KolliRT,BentleyKE,RushmoreJ,WongK,ShawTI,RothrockMJ,Jr..,McKeeAM,GuoTL,MauricioR,MolinaM,CummingsBS,LashLH,LuK,GilbertGS,HubbellSP,FairclothBC.2019.AdapteramaII:universalampliconsequencingonIlluminaplatforms(TaggiMatrix).PeerJ7:e7786.OpenUrl35.↵LawrenceCB,SolovyevVV.1994.Assignmentofposition-specificerrorprobabilitytoprimaryDNAsequencedata.NucleicAcidsRes22:1272–80.OpenUrlCrossRefPubMed36.↵CaporasoJG,LauberCL,WaltersWA,Berg-LyonsD,HuntleyJ,FiererN,OwensSM,BetleyJ,FraserL,BauerM,GormleyN,GilbertJA,SmithG,KnightR.2012.Ultra-high-throughputmicrobialcommunityanalysisontheIlluminaHiSeqandMiSeqplatforms.ISMEJ6:1621–4.OpenUrlCrossRefPubMedWebofScience37.↵WuWK,ChenCC,PanyodS,ChenRA,WuMS,SheenLY,ChangSC.2019.Optimizationoffecalsampleprocessingformicrobiomestudy-Thejourneyfrombathroomtobench.JFormosMedAssoc118:545–555.OpenUrl38.↵KlindworthA,PruesseE,SchweerT,PepliesJ,QuastC,HornM,GlocknerFO.2013.Evaluationofgeneral16SribosomalRNAgenePCRprimersforclassicalandnext-generationsequencing-baseddiversitystudies.NucleicAcidsRes41:e1.OpenUrlCrossRefPubMed39.↵GhyselinckJ,PfeifferS,HeylenK,SessitschA,DeVosP.2013.Theeffectofprimerchoiceandshortreadsequencesontheoutcomeof16SrRNAgenebaseddiversitystudies.PLoSOne8:e71360.OpenUrlCrossRefPubMed40.↵RCoreTeam.2015.R:Alanguageandenvironmentforstatisticalcomputing,RFoundationforStatisticalComputing,Vienna,Austria,https://www.R-project.org/.41.↵OksanenJ,BlanchetFG,KindtR,LegendreP,MinchinPR,O’HaraRB,SimpsonGL,SolymosP,StevensMHH,WagnerH.2015.vegan:communityecologypackage.Rpackageversion2.3-1,Oulu,Finland.https://CRAN.R-project.org/package=vegan. Backtotop PreviousNext Posted August20,2020. DownloadPDF Email ThankyouforyourinterestinspreadingthewordaboutbioRxiv.NOTE:Youremailaddressisrequestedsolelytoidentifyyouasthesenderofthisarticle. YourEmail* YourName* SendTo* Entermultipleaddressesonseparatelinesorseparatethemwithcommas. Youaregoingtoemailthefollowing EvaluationofCompatibilityof16SrRNAV3V4andV4AmpliconLibrariesforClinicalMicrobiomeProfiling MessageSubject (YourName)hasforwardedapagetoyoufrombioRxiv MessageBody (YourName)thoughtyouwouldliketoseethispagefromthebioRxivwebsite. YourPersonalMessage CAPTCHAThisquestionisfortestingwhetherornotyouareahumanvisitorandtopreventautomatedspamsubmissions. Share EvaluationofCompatibilityof16SrRNAV3V4andV4AmpliconLibrariesforClinicalMicrobiomeProfiling Po-YuLiu,Wei-KaiWu,Chieh-ChangChen,SuraphanPanyod,Lee-YanSheen,Ming-ShiangWu bioRxiv2020.08.18.256818;doi:https://doi.org/10.1101/2020.08.18.256818 ShareThisArticle: Copy CitationTools EvaluationofCompatibilityof16SrRNAV3V4andV4AmpliconLibrariesforClinicalMicrobiomeProfiling Po-YuLiu,Wei-KaiWu,Chieh-ChangChen,SuraphanPanyod,Lee-YanSheen,Ming-ShiangWu bioRxiv2020.08.18.256818;doi:https://doi.org/10.1101/2020.08.18.256818 CitationManagerFormats BibTeXBookendsEasyBibEndNote(tagged)EndNote8(xml)MedlarsMendeleyPapersRefWorksTaggedRefManagerRISZotero TweetWidgetFacebookLikeGooglePlusOne SubjectArea Microbiology SubjectAreas AllArticles AnimalBehaviorandCognition(3579) Biochemistry(7526) Bioengineering(5486) Bioinformatics(20703) Biophysics(10261) CancerBiology(7939) CellBiology(11585) ClinicalTrials(138) DevelopmentalBiology(6574) Ecology(10145) Epidemiology(2065) EvolutionaryBiology(13556) Genetics(9502) Genomics(12796) Immunology(7888) Microbiology(19460) MolecularBiology(7618) Neuroscience(41917) Paleontology(307) Pathology(1253) PharmacologyandToxicology(2182) Physiology(3253) PlantBiology(7011) ScientificCommunicationandEducation(1291) SyntheticBiology(1942) SystemsBiology(5410) Zoology(1108)