Region-specific similarity cutoff for bacterial species circumscription in 16S rRNA-based studies

Abstract

16SribosomalRNAisthemostwidelyusedmolecularmarkerinmicrobialcommunitystudies.Generally,16SrRNAgenesequencesareclusteredintooperationaltaxonomicunits(OTUs)basedonsequencesimilarityandtheOTUdefinedat97%similaritycutoffiscommonlyusedasa'proxy'forbacterialspecies.Theproblemisthatthecutofforiginallysetforthefulllength16SrRNAhasalsobeenusedinsubregionsof16SrRNAwithoutpropervalidation.Here,wetestedfirstthesuitabilityofusingsequencesimilarityitselfinthreesubregionsof16SrRNAandthendeterminedtheoptimalsimilaritycutoffforspeciesdemarcationineachregiononthebasisofgenomicevidence.Theeffectofalignmentanddistancecalculationmethodsonsimilaritycutoffvalueswasadditionallyinvestigated.F-measureanalysisrevealedthatV1-V3regionshowedthehighesttaxonomicaccuracy,asimilarextenttothatoffulllength,andloweredinothertworegions,V3-V5andV6-V9regions.Optimalsimilaritycutoffcorrespondingtobacterialspeciesboundarydifferedbyregions.SimilardegreeofcutoffwasobtainedforV3-V5(99.0%)andV6-V9regions(99.1%),whereasamuchlowerlevelofcutoffwasestimatedforV1-V3region(98.0%).Alignmentmethodshavelittleinfluenceontheoptimalcutoffvalues,whilethecutoffvaluesdifferedmorebydistancecalculationmethods.Takentogether,ourresultsshowthatregion-optimizedsimilaritycutoffshouldbeusedwhenclusteringsubregionsof16SrRNAintoOTUsandamongthreesubregionsV1-V3isthemostappropriatetargetforsimilarity-basedOTUclustering.