Extended cone-curvature based salient points detection and 3D model retrieval20
上亿文档资料，等你来发现
Extended cone-curvature based salient points detection and 3D model retrieval20
MultimedToolsAppl(2013)6；Extendedcone-curvatureba；YuJieLiu&XiaoDongZhang&Z；Publishedonline:3January；#SpringerScience+Busines；AbstractLocalfeatureextr；Y.Liu(*):Z.Li；SchoolofComputer
MultimedToolsAppl(C693DOI10.-011-0950-7Extendedcone-curvaturebasedsalientpointsdetectionand3DmodelretrievalYuJieLiu&XiaoDongZhang&ZongMinLi&HuaLiPublishedonline:3January2012#SpringerScience+BusinessMedia,LLC2011<paredwiththeglobalfeature,itismoresuitabletodothepartialretrievalandmorerobusttothemodeldeformation.Inthispaper,alocalfeaturecalledextendedcone-curvaturefeatureisproposedtodescribethelocalshapefeatureof3Dmodelmesh.Basedontheextendedcone-curvaturefeature,salientpointsandsalientregionsareextractedbyusinganewsalientpointdetectionmethod.Thenextendedcone-curvaturefeatureandlocalshapedistributionfeaturecalculatedonthesalientregionsareusedtogetherasshapeindex,andtheearthmover’sdistanceisemployedtoaccomplishsimilaritymeasure.Aftermanytimes’retrievalexperiments,thenewextendedcone-curvaturedescriptorweproposehasmoreefficientandeffectiveperformancethanshapedistributiondescriptorandlightfielddescriptorespeciallyondeformablemodelretrieval.Keywords3Dmodelretrieval.Extendedcone-curvature.Salientpoints.Earthmover’sdistanceY.Liu(*):Z.LiSchoolofComputerScienceandCommunicationEngineering,ChinaUniversityofPetroleum,Qingdao266555,Chinae-mail:bilin_ResearchHighlightsExtendsthecone-curvatureanddefinetheconceptofextendedcone-curvatureofgeneralmeshandgivesamethodtocomputeextendedcone-curvatureontriangularmeshes.Analgorithmisproposedtodetectthesalientpointsbasedonextendedcone-curvatureandsalientregionsareextractedaccordingtothesesalientpoints.Extendedcone-curvaturefeatureandshapedistributionfeatureofthesalientregionsareusedtogetherasthemodelindexwhichperformsverywellinthe3DmodelretrievalwithEarthMover’sDistanceasthesimilaritymeasure.X.ZhangCenterforHumanComputerInteraction,ShenzhenInstituteofAdvancedIntegrationTechnology,Qingdao266555,ChinaH.LiNationalResearchCenterforIntelligentComputingSystems,InstituteofComputingTechnology,ChineseAcademyofSciences,Beijing100190,China672MultimedToolsAppl(C6931IntroductionSincethefastdevelopmentof3Dtechnologiesandcomputergraphicsattheendofthetwentiethcentury,3Dmodelhasbeenwidelyused,suchascomputergames,movies,mechanicalCAD,virtualrealityandsoon.Atthesametime,theInternetprovidesanotherplatformtofacilitateitswideapplicationTherefore,theamountof3Dmodelsbecomeslargerandlarger.3Dmodelhasbeenregardedasthefourth-generationmultimediainfor-mationaftervoice,imageandvideo.However,3Dmodelingwithahighprecisionisstilladifficultandtime-consumingprocedure.Ifwecanreusetheexistingmodels,itwillbringagreatimprovementofthemodelingefficiency.Buthowcanwefindtheexactmodelsfrommillionsof3dmodelswithintheshortesttime?Keyword-based3Dmodelretrievalhasbeenusedtoassistuserstofinddesiredmodelsinitially.Everymodelin3Dmodeldatabaseislabeledbyoneormorekeywords.Althoughthismethodhashighretrievalspeedandaccurateresults,twodisadvantagespreventitfrombeingwidelyapplied.Firstly,it’saverytime-consumingandlaborioustasktolabelall3dmodelsifthedatabaseishuge.Secondly,thekeywordscanbeeasilyaffectedbyhumans’subjectivethoughts.Differentpeoplemayhavedifferentconceptsforthesamemodel.Sothistraditionalmethodcannotmeetallusers’demands.Thecontent-based3Dmodelretrievalhasbeenaresearchhotspotsince2000.Itbreaksthroughtheconstraintsofthekeyword-basedmethodandmakesuseofthevisualfeatureofthemodelsasthemodelindexdirectly.Atpresentthereareseveralcontent-based3Dmodelretrievalsystemsforinstance,thePrincetonUniversity’s3Dmodelretrievalsystem[6].Therearefivemainmodulesintheseretrievalsystems:theuserinterface,3Dmodeldatabase,featuredatabase,featureextractionandsimilaritycomputation.ThestructureofthesystemisshowninFig.1.Atfirst,featuresofallthemodelsinthemodeldatabaseareextractedofflinebyusingfeatureextractionalgorithmandthensavedinthefeaturedatabase.Whenusersubmitsaquerymodelthroughtheuserinterface,thefeatureofthequerymodeliscalculatedsoon.Thenthesimilaritybetweenthequerymodelandallmodelsinthedatabasearecomputedbythedistancebetweenthefeatureofthequerymodelandthefeaturesinthefeaturedatabase.Afterthat,thedatabasemodelsarerankedbythedistance.Themodelsrankedinthemosttoparethemostsimilartothequerymodel.Theranklistisreturnedgraphicallytotheuserinterfacewheretheretrievalresultisdisplayed.Featureextractionof3Dmodelsdefinitelyplaysaveryimportantroleintheretrievalsystem.Althoughtherehavebeentremendousfeatureextractionmethods,noneofthemcanbefitforallsituations.Manyfeaturedescriptorsfromthesemethodsusuallyarenon-topologicalfeaturesthatcanbeaffectedbyshapedeformationeasily.Sometimestopologicalfeaturessuchasgraphorskeletonarehiredtosolvetheproblem,buttheyneedmorecomputationandareverysensitivetomodelnoisewhichwillbeanalyzedintherelatedworksection.Inthispaperalocalshapefeatureextractionmethodcalledextendedcone-curvatureisproposed.Therestofthispaperisorganizedasfollows.Therelatedworkisdescribedinsection2.Insection3extendedcone-curvatureisdefinedandthealgorithmtocomputeextendedcone-curvatureofeachtriangularmeshisalsoexplained.Section4givesthealgorithmtodetectthesalientpointsandsalientregionsof3DmodelsandSection5introducestheearthmover’sdistancetocomputesimilaritybetweenthesesalientfeatures.Thentheexperimentalresultsareshowninthesection6.Finally,conclusionsaregiveninsection7.MultimedToolsAppl(C6936733D Model Database Feature Extraction Query Model Feature Database Similarity Computation Model Rank List User Interface Fig.1Thestructureof3Dmodelretrievalsystem2Relatedwork2.1GlobalfeatureextractionmethodsAtpresent,manyfeatureextractionmethodshavebeenproposedtopreciselydescribe3Dmodels.Generally,thesemethodscanbedividedintothreecategories:thegeometry-basedmethods,thetopology-basedmethodsandtheview-basedmethods.Geometry-basedmethodsmainlyanalyzethegeometrypropertiesof3Dmodels.Statis-ticalstrategiesareveryoftenemployedbythesegeometry-basedmethods.HornproposedtheextendedGaussianimagemethodtodescribethe3Dmodel[8].Itsprincipleistocountthenormalvectordistributionaccordingtotheirdirections.Cord-basedmethodwaspro-posedtodescribetherelationshipbetweenthemodelvertexandprincipalaxis[13].Firstly,theprincipalaxesarecomputedthroughprincipalcomponentanalysis.Thentheanglebetweenmodelverticesandthefirstprincipalaxis,theanglebetweenmodelverticesandthesecondprincipalaxisareallcountedtoobtainthefeaturevector.Onemoremethodusingstatisticalstrategy,liketheshapehistogramproposedin1999mainlyanalyzesthespatialdistributionofmodelvertices[4].Threepartitionmethodsaregiventodividethemodelintoseveralparts.Thefeaturevectoriscomputedbycountingthevertexnumberofeverypart.Onedisadvantageofthesemethodsisthatthefeaturevectorisnotinvarianttodifferentmodelresolutions.Afterwards,Osadaetal.[12]proposedamethodusingtheprobabilitydistributionofgeometrypropertiescomputedfromrandomverticessampledonthemodelsurfaceasadescriptor,whichiscalledshapedistribution.Fiveshapefunctionsareappliedtocomputethegeometrypropertiesincludingarea,angle,distanceandvolume.Amongthem,D2whichmeasuresthedistancebetweenanypairofrandomverticesonthesurfaceof3Dmodeldescribesthemodelfeaturebest.Shapedistributionisinvarianttothemodeltranslation,rotation,resolutionandnoise.Anditiseasytocompute.Butitstillcannotsolvetheproblemofmodeldeformation.Topology-basedmethodusuallyrepresentsthemodelbytreeorgraphstructuresothatthemodelstructurecanbewelldescribed.Amentausedskeletontreecalculatedbyvoronoigraphtorepresentmodels[3].Thenthesimilarityofmodelscanbecomputedbymatching674MultimedToolsAppl(C693theseskeletontrees.Hilagaetal.proposedamethodcallmulti-resolutionreebgraph(MRG)[7].Byusingthegeodesicdistanceasthemorsefunction,everymodelcanberepresentedasseveralreebgraphindifferentresolutions.Thenthesimilarityofmodelscanbeobtainedbycomparingtheircorrespondingreebgraphs.GaryK.L.Tametal.extractedthetopologicalpointsandtopologicalringstorepresentmodels[18].Topologicalfeatureandgeometricfeaturearejoinedtogethertocomputethesimilarityofthemodels.Themainadvantageoftopology-basedmethodsisthattheyarerobusttothemodeltranslation,rotation,scalingandespeciallytomodeldeformationthatmostoftheothermethodscannotsolvewell.However,extractionofmodelskeletonorgraphistime-consumingandverysensitivetothefinecomponentsofmodels.Inaddition,graphmatchingisaNP-completeproblemwithlargecomputation.View-basedmethodisalsoanimportantcategoryamongfeatureextractionmethods.Theyareinlinewithhuman’svisualperceptionandrobusttomodelnoise,simplification,andrefinement.Ohbuchietal.proposedthedepthbufferimagemethodtoextractmodelfeature[11].42viewpointslocatedonthemodel’sboundingboxareusedtoobtain42depthbufferimagesofthemodel.ThenthegenericFourierdescriptorsareextractedfromtheseimagesastheindexofthemodel.ThesimilarityofmodelscanbeobtainedbycomputingthesegenericFourierdescriptors.Chenproposedthelightfielddescriptormethod[5].Similartodepthbufferimage,ituses20viewpointslocatedontheboundingsphereofthemodeltoobtaintheprojectionimages.ThenthezernikemomentsfeaturesarecalculatedfromtheseimagesandtheFourierdescriptorsareextractedfromthecontouroftheseimages.TheprincipalplanewasfirstusedinthemodelfeatureextractionbyChen-TsungKuoa[9].Aftertheprincipalplanewascomputed,everyvertexofthemodelcanbeprojectedontheplane.Accordingtothedistributionoftheseprojectionvertices,featurevectorisextracted.Althoughview-basedmethodsareeasytounderstandandcompute,theycannotdealwithmodeldeformation.That’sbecauseprojectionofthedeformedmodelismostlydifferentfromthatoftheoriginalmodel.2.2LocalfeatureextractionmethodsAllthemethodsmentionedaboveareglobalfeaturebasedmethodsdescribingthemodelfromtheviewofthewholemodelshape.Localfeatureshavebeenpaidmoreattentionbecforexampletheycanbeusedinpartialretrieval.Inrecentyearsseverallocalfeatureextractionmethodshavebeenproposed.Buthowtodefinethelocalregionisstillaproblemunsolved.PhilipShilaneetal.proposedamethodusingthedistinctionoflocalmeshtoextractsalientfeature[16,17].Thelocalregionisdefinedasthesurfaceregionboundedbyaspherecenteredonverticessampledfrommodelsurfacewithoneradius,whichisproportionaltotheradiusofthemodel(forexample0.5).Whentheselocalregionsaredefined,harmonicshapedescriptorsareextractedasthelocalfeaturesareusedtodoretrieval.Theretrievalresultsareevaluatedbydiscountedcumulativegainwhichisdefinedasthedistinctionofthelocalregion.Thelocalfeaturesoftheregionwiththegreatestdistinctionareselectedasthesalientfeaturestoindexthemodel.Thismethodtriedtofindthelocalregions,whicharepartialsimilartomodelsinthesameclasswhiledifferentfrommodelsintheotherclasses.Itsdisadvantageisthattheradiusofthesphereishardtoget.Andtheregionboundedinthesphereislikelytobeseveralpartsthatarenotconnected.Inaddition,hugeamountofcomputationcosthastobeconsideredoncalculatingthedistinctionofthelocalregion.LiuYietal.proposedapartialretrievalmethod[10].Atfirst,spinimagefeaturesareextractedfromallthemodelsinthedatabase.Bymakinguseofthebagofwordsalgorithm,MultimedToolsAppl(C693675allthesefeaturesareusedtoclustertogetsomevisualwords.Theneverymodelcanberepresentedbyawordfrequencyhistogram.KullbackCLeiblerdivergenceisselectedtodothepartialsimilaritymeasure.Itsavesalotofspacebecauseitdoesnotneedtosaveallthesespinimagefeatures.Andtherepresentationoffeatureissimple,whichalsomakesthesimilaritymeasuremucheasier.However,bytransformingthesespinimagefeaturestoafrequencyhistogram,somespaceshapeinformationwouldbelost.Furthermore,whenmodel’sdatabasechanged,thevisualwordsneedtobecomputedonceagain.S.Shalometal.gaveamodelsegmentationmethodtoextractmodellocalfeature[15].Basedontheshapediameterfunction,everymodelcanbesegmentedintoseveralcompo-nentsatdifferentlevels.Sothemodelcanberepresentedbyaleveltree.Thedeeperintheleveltreethemodellocates,themorerefinedthemodelissegmented.Aftersegmentation,localfeaturesareextractedfromthesecomponents.Notonlylocalgeometricfeature,butalsothecontextofthelocalfeatureintheleveltreeisconsideredtocomputethesimilarityinpartialretrieval.Ithasgoodretrievalresults.Butitdependsonthemodelsegmentationresults.AntonioAdánmadeuseofCone-Curvatureasthelocalfeaturetocomputesimilaritiesbetweenmodels[1].EverymodelneedstobeapproximatedbyaunitsphereshowninFig.2.ConeCurvatureisdefinedonthedeformedspherearoundeveryvertex.Thenthecone-curvaturematrixcomposedofallcone-curvaturesisusedtoindexthemodelandcomputethesimilarity.Theauthorextendedthecone-curvaturetostandardtriangularmeshtoclustermodels[2].Howeverthemodelsmustberegularizedandresampledtoafixednumberofnodes.ThenModelingWave(MW)usedtocomputecone-curvatureorganizestherestofthenodesofthemeshinconcentricgroupsspatiallydisposedaroundsomesourcepoint.Butthepreprocess-ingstepsuchasregularizationandresamplemadeitnotsoadaptableandnoteasytocompute.Soweextendthecone-curvaturetocomputedirectlyontriangularmeshmodels.Basedonitsalientfeaturesareextractedandusedtoretrieve3Dmodels.Inthispaperweproposeanovellocaldescriptorextractionmethod.Thenoveltyofthispaperisthree-fold.Atfirstweextendthecone-curvatureandproposeanalgorithmtocomputeitonthetrSecondbasedontheextendedcone-curvatureasalientpointdetectionmethodisproposedtoextractthesalientlocalfeatureof3DThirdtheearthmover’sdistanceisemployedtocomputethesimilarityofmodelswhichareindexedbyasetofsalientfeatures.Detailsareexplainedasthefollowing.3Extendedcone-curvatureAsmentionedbeforethecone-curvatureproposedin[1,2]cannotbeuseddirectlyonstandard3Dmodelswhichareusuallytriangularmeshes.ItneedsapreprocessingstepFig.2Definitionofcone-curvature三亿文库包含各类专业文献、文学作品欣赏、生活休闲娱乐、中学教育、行业资料、应用写作文书、高等教育、幼儿教育、小学教育、Extended cone-curvature based salient points detection and 3D model retrieval20等内容。　
　 点击 use curvature 和 use proximity 选择单元体...Washer split:快速建立垫圈 Add/removed points:添加...cone 中,选 full cylinder, bottom center 选圆柱... 　Point Polar: 在球面坐标上创建点 Point XYZ: 通过... 测试一条曲线是闭合的还是周期性的 Curvature: ... 在平面的中心点创建立方体 Cone: 圆锥体 Cylinder:...fulltext基因组学2 三亿文库
fulltext基因组学2
ApplMicrobiolBiotechnolDOI10.-011-3520-zMETHODSANDPROTOCOLSValidationoftheH2SmethodtodetectbacteriaoffecaloriginbyculturedandmolecularmethodsLanakilaMcMahan&AnthonyA.Devine&AmyM.Grunden&MarkD.SobseyReceived:10May2011/Revised:7June2011/Accepted:28July2011#Springer-Verlag2011AbstractUsingbiochemicalandmolecularmethods,thisresearchdeterminedwhetherornottheH2Stestdidcorrectlyidentifysewage-contaminatedwatersbybeingthefirsttouseculturingandmolecularmethodstoidentifythetypesandnumbersoffecalindicatororganisms,pathogens,andothermicrobespresentinsewagesampleswithpositiveH2Stestresults.Fortheculture-basedmethod,sampleswereanalyzedforthepresenceoffecalbacteriabyspreadplatingthesewagesampleontodifferentialandselectivemediaforAeromonasspp.,Escherichiacoli,sulfite-reducingclostridia,H2S-producingbacteria,andSalmonella/Shigellaspp.Theisolateswerethen:(1)testedtodeterminewhethertheywereH2S-producingorganismsand(2)identifiedtothegenusandspecieslevelusingbiochemicalmethods.Themolecularmethodusedtocharacterizethemicrobialpopulationsofselectsampleswasterminalrestrictionfragmentlengthpolymorphisms.TheseexperimentsonsewageprovidedevidencethatpositiveH2Stestsconsistentlycontainedfecalbacteriaandpathogens.TherewerestrongrelationshipsofagreementbetweentheorganismsidentifiedbybothL.McMahan(*):M.D.SobseyDepartmentofEnvironmentalSciencesandEngineering,SchoolofPublicHealth,UniversityofNorthCarolinaatChapelHill,ChapelHill,NC27599,USAe-mail:kmcmahan@email.unc.eduA.M.GrundenDepartmentofMicrobiology,NorthCarolinaStateUniversity,Raleigh,NC27695,USAA.A.DevineCenterforMicrobialCommunitySystemsandHealthResearch,RTIInternational,ResearchTrianglePark,ChapelHill,NC27709,USAe-mail:methodstested.Thisstudyisanimportantadvanceinmicrobialwaterqualitydetectionsinceitisfocusedontheevaluationofanovel,low-cost,watermicrobiologytestthathasthepotentialtoprovidemillionsofpeopleworldwideaccesstowaterqualitydetectiontechnology.Ofprimeconsiderationinevaluatingwaterqualitytestsisthedetermi-nationofthetest’saccuracyandspecificity,andthisarticleisafundamentalstepinprovidingthatinformation.KeywordsH2Stest.TRFLP.MicrobialwaterqualityIntroductionThereisgreatandurgentneedforarapid,simple,inexpensivetesttodeterminethelevelsoffecalcontaminationinwaterinresource-poorsettings.Inthedevelopingworld,lackofaccesstolaboratoriesorfieldanalysiskitsisanobstacletotheidentificationandthereforeprovisionofmicrobiologicallysafedrinkingwaterformanycommunitiesandpeople(SobseyandPfaender2002).Thehydrogensulfide(H2S)bacteriatestwasdevelopedasalow-costalternativefecalindicatortestthatdoesnotrequireamicrobiologylaboratoryorabacteriologicalfieldtestkit(ManjaandKaul1992).Thetestisintendedtodetectthosebacteriaoffecaloriginthatareabletoreduceorganicsulfurtosulfide(asH2Sgas),whichthenreactsrapidlywithirontoformablackprecipitate(Widdel1988).TheadvantagesofthistestincludetheeaseofculturinganddetectingH2S-producingmicroorganismsandthesimplicityandlowcostofapresenceCabsenceH2Stestformatforthesemicroorganisms.VariousinvestigatorshaveexaminedtheH2Smethodanditsmodificationsintropicalandtemperateregionsinavarietyofwatertypes(Manjaet.al1982;Rattoet.al1989;Castilloetal.1994;Anwaretal.1999;GentheandFranck1999;PathakandGopal2005;Roseretal.2005;Guptaetal.2008;Rijaletal.2001)andcomparedittotraditionalbacterialindicatorsoffecalcontaminationinwater.OneofthemajorweaknessesoftheH2Stestforthedetectionoffecalbacteriaisthevariationinsensitivityandspecificityforbacteriaoffecaloriginobtainedacrossdifferentstudies.PreviousstudiesapplyingtheH2Stesttogroundwatersampleshavedemonstratedfalsepositiveresults,whereH2S-positivesamplescontainednofecalcoliformsorEscherichiacoli(Pantetal.2002;Kasparetal.1992).Falsenegativeresults,whereH2S-negativesampleswerefoundtocontainE.coli,havebeenreportedinotherstudies(Tewarietal.2003).FortheH2Sbacteriatesttobeanacceptabletooltoevaluatewaterqualityforthepresenceandmagnitudeoffecalcontamination,dataareneededindicatingwhichmicroorgan-ismsproducepositiveresultsinthetestandunderwhatconditionstestresultsindicateactualfecalultimately,aquantitativeversionofthetestisneededtoestimatethemagnitudeoffecalcontamination.ThepurposeofthisresearchwastovalidatetheH2Stestusingcultureandmolecularmethods.Forthetesttobeeffective,itmustfirstbecapableofdetectingfecalbacteriawhenappliedtosewagesamplesaswaterscontaminatedwithhumansewageareusedasdrinkingwatersourcesinmanyimpoverishedareasworldwide.Thisisthefirstresearchtouseculture-basedbiochemicalandculture-independentmoleculartechniquestodeterminethetypesofmicrobialcommunitymembers,includingfecalindicatororganisms,pathogens,andothermicrobespresentinhumansewagesamplesthataredetectedinaquantitativeH2Stestasmicroorganismsofconcerntohumanhealth.Moleculargeneticstechniquesutilizingextractednucleicacidsnowallowmicrobialcommunityanalysistobecoupledwithaphylogeneticframework.Terminalrestric-tionfragmentlengthpolymorphism(TRFLP)wasthenucleicacid-basedmethodemployedinthisstudybecauseitprovidesawaytodeterminethepresenceofcommonspeciesinasamplewithorwithoutculturingtheorganisms,facilitatesfindingmajordifferencesbetweencommunities,andallowsfortestinghypothesesbasedonacomparisonofsamples(Kentetal.2003).ByusingTRFLP,Liuetal.(1997)wereabletodistinguishallbacterialspeciesinamodelbacterialcommunity,andthepatternwasconsistentwiththepredictedoutcome.OthershavecomparedtheresultsofTRFLPandthetraditionalculture-basedapproachandfoundthatTRFLPoftenprovidesamoredetailedanalysisthanthetraditionalapproach(Pidiyaretal.2004;Moralesetal.2006).Inthisstudy,theresultsofTRFLPandtraditionalculture-basedisolationandbiochemicalcharac-terizationmethodswerecomparedtodeterminewhatmicroorganismsaregrowinginpositiveH2Ssamplesfrommunicipalsewage.IfbothtechniquesfoundthatthisnewH2StestwaseffectivewithsamplesthatshouldhaveApplMicrobiolBiotechnolorganismsofconcern,futureanalysiscanfocusontheeffectivenessofthetestwithwatersthatmayormaynotcontainorganismsnormallyfoundinsewageandotherfecalwastesourcesthatareaconcerntohumanhealth.MaterialsandmethodsCulture-basedbiochemicaldetectionmethodsTodeterminethegeneraandrelativenumbersofbacteriapresentinsewage-spikedphosphate-bufferedsaline(PBS)samples,grabsamplesof120-mLvolumesofrawuntreatedsewagewereobtainedonthreeseparateoccasionsfromtheOrangeWaterandSewerAuthority(OWASA)wastewatertreatmentplant(ChapelHill,NC).TheOWASAsystemservesauniversitycommunityhavingnomajorsourcesofindustrialwastes.Eachsampleofcollectedsewagewasconsideredaseparateexperimentsincesampleswerecollectedevery2weeksovera6-weekinterval.Fromthe120-mLraw,untreatedsewagesample,duplicate10mLaliquotswereremovedandpelletedat3,500rpmfor20min,andthepelletswereoverlaidwith300μLoftheoriginalsampleandarchivedat?80°CforfutureDNAanalysis(seeFig.1).PathoscreenreagentforH2Sbacteriatestingofa100-mLwatersample(HachCompany,Loveland,CO)wasaddedtotheremaining100mLofrawsewageinafive-compartmentMPNbagandincubatedat37°Cfor24h.TheMPNbagisaclearpolyethylenebag,15cmwide×23cmlong(Whirl-Pak?,Nasco,FortAtkinson,WI)inwhichtherearefiveinternalverticalcompartmentsinthelowertwothirdsofthebag,eachwithavolumeof1,3,10,30,and56mL,respectively.Thefive-compartmentbaggaveanMPNestimateofH2S-producingorganisms.Forbiochemicalanalysisofbacteriafromtheculture-baseddetectionmethod,a0.5-mLsamplefromeachH2S-positivecompartmentofanH2SMPNbagsamplewasdilutedseriallytenfoldinPBStoadilutionof10?5.Then,100μLvolumesofeachdilutionwerespreadontoduplicate13×150-mmdiameterplatesofthefollowingagarmediatoisolatecolonies:Bio-RadRAPID'E.coli2agar,SalmonellaCShigellaagar,Phenyl-ethanolagar,m-Aeromonasselectiveagar,andH2Sagar(Rijaletal.2001).Allplateswereincubatedaerobicallyat37°Cfor24h.IsolatesfromspreadplateswereobtainedbystreakplatingcharacteristiccoloniesontoTrypticsoyagaronthreesuccessivedays.Thesecolonyisolateswerearchivedin0.8mLofTrypticsoybrothat?80°C.TheisolateswerethentestedtodeterminewhethertheyproduceH2SbyculturinginH2Smediumandwerefurtheridentifiedatthegenusandspecieslevelsusingstandardbiochemicalidentificationtestkits,specificallyBBLEnterotubeII(BDDiagnosticSystems,Sparks,MD)andtheAPI20ESystem(bioMerieux,Inc.,Hazelwood,MO).ApplMicrobiolBiotechnolFig.1Workflowdiagramdescribingoftheculture-basedbiochemicalidentificationandtheTRFLPprocessesDNAextractionTwocompartmentbagsofpositiveH2Smediawerepelleted,andtheresultingpelletswereoverlaidwith1mLoftheH2S-positivesamplespentmediumandarchivedat?80°CforfutureDNAisolationforTRFLPmolecularcommunityanalysis(Fig.1).GenomicDNA(gDNA)wasextractedusingtheMoBioPowersoilTMDNAextractionkit(MOBIOLaboratoriesInc.,SolanaBeach,CA)accordingtothemanufacturer’sprotocol.AgarosegelelectrophoresiswasusedtovisualizewhethersufficientqualitygDNAwasisolatedfromeachsample.A3-μLvolumeofeachgDNAsamplewaselectrophoresedthrougha1%TAEagarosegelcontaining15μLofethidiumbromideper100mLofagarosegel.IsolatedgDNAwasstoredat?80°CuntilitwasusedforPCRreactions.PCRconditionsThree-microlitervolumesofeachDNAsamplewereaddedto97μLofMasterMix[persample:10μL10XReactionB0.8μLdNTP;83.7μLPCR0.5μLeachofthebacterial-specific16SrDNAprimers8F-Hex5′-AGAGTTTGATC(A/C)TGGCTCAGandreverseprimer1492R5′-GGTTACCTTGTTACGACTT;0.5μLofQiagenHotstarTaqDNApolymerase(Qiagen,Valencia,CA)].EachDNAsamplewasamplifiedintriplicate.TheforwardprimerforthePCRreactionwaslabeledonthe5′endwithahexamidefluorescentmarkertoallowtheterminalfragmenttobetracked.PCRwasperformedinaPerkin-Elmer9600thermocyclerusinganinitialdenatur-ationstepof15minat95°C,followedby35cyclesconsistingofdenaturation(1minat94°C),annealing(1minat50°C),andextension(2minat72°C)andafinalextensionat72°Cfor7min.PCRreplicatesofeachsamplewerethenpooledandpurifiedusingtheUltraCleanTMPCRClean-upKit(MOBIOLaboratoriesInc.)accordingtothemanufacturer’sprotocol.TerminalrestrictionfragmentlengthpolymorphismFortheTRFLPanalysisoftheamplifiedbacterial16SrDNAs,threerestrictionenzymes,RsaI,HhaI,andMspI(NewEnglandBiolabs,Inc.,Ipswich,MA),wereused.FortheRsaIdigest,30μLofpurifiedPCRproduct(approx-imately30μg)wasmixedwith10μLofReactionBuffer1,59μLPCRgradewater,and1μLofrestrictionenzyme.FortheHhaIdigest,30μLofpurifiedPCRproduct(approximately30μg)wasmixedwith10μLofReactionBuffer4,1μLbovineserumalbumin(BSA),58μLPCRgradewater,and1μLofrestrictionenzyme.ForMspI,30μLofpurifiedPCRproduct(approximately30μg)wasmixedwith10μLofReactionBuffer4,1μLBSA,58μLPCRgradewater,and1μLofrestrictionenzyme.Restrictiondigestswereincubatedovernightat37°C.Forcleanup,restrictiondigestswereheat-treatedat60°Cfor20mintoheat-inactivatetherestrictionenzymes.TheQIAquickNucleotideRemovalKit(Qiagen,Hilden,Germany)wasthenusedtopurifythedigestedDNAaccordingtothemanufacturer’sprotocol,except50μLofwarmed(60°C)PCRgradewaterwasaddedinsteadofkitelutionbuffer,andthewaterwasallowedtoincubateonthecolumnfor5minpriortoelutionoftheDNA.DNAsampleswerethenfrozenat?20°C.FragmentdetectionwascarriedoutattheMSUGenomicsTechnicalSupportFacilityaccordingtotheirdetectionprotocols(http://gtsf.msu.edu/dna-fingerprinting-and-genotyping).FragmentanalysisDatatablescontainingfragmentsizeandabundancedataforeachdigestoftheDNAofsewagesampleswereexportedfromGeneScan,andtheresultingtextfilesweresentforpatterndetectionbytheInSilico?database.Eachfilecontainedallthedetectedfragmentsforagivenrestrictiondigest(e.g.,dataobtainedfromoneoftheHhaIdigestsforasamplewouldbecontainedinonefile,MspIfragmentswouldbecontainedinanotherfile,andRsaIfragmentdatawouldbecontainedinathirdfile).Eachentryinthesedatafilescontainedfragmentlengthsize,retentiontimeongel,peakheight,fragmentidentificationnumber,andapeakareafoundinthesample.Forcalculationofthediversityindices,theTRFLPanalysispeakareawasusedastheamountmeasurement,anditsrelativeabundancewasmeasuredbydividingindividualpeaksbythetotalfluorescenceofthesample.Theresultsforeachdiversitymeasurearerepresentativeofthenumberoffragmentsineachexperimentalsample.InSilico?outputPatterndetectionandpatternidentificationwerecarriedoutusingtheInSilico?softwarepackage(InSilico?,RTP,NC).TheInSilico?outputisacomprehensivedatasetwhichincludesthefollowingdiversityanalysisvalues:Simpsonindexofdiversity,ameasureoftherichness(thenumberofdifferentspeciespersample)reciprocalSimpson,aninverseoftheSimpson’sindex(lowestvalueis1,thehigherthevaluethegreaterthediversity);speciesrichness(thenumberofspecieswithinacommunity);andtheShannonCWeaverdiversityindex,whichisoneofseveraldiversityindicesusedtomeasurediversityincategoricaldataandtakesintoaccountthenumberofspeciesandtheevennessofthespecies.Theindexisincreasedeitherbyhavingadditionaluniquespeciesorbyhavinggreaterspeciesevenness(KimandMarsh2004;Blackwoodetal.2007).InSilico?alsoApplMicrobiolBiotechnolprovidesinformationonthefragmentparameters,including:totalfragmentutilization,whichistheproportionofthefragmentsusedintheanalysiscomparedtothetotalnumberthetotalnumberoffragandevenness,whichisameasuredescribinghowmucheachindividualfragmentcontributestothewhole(ona0C1scale,closerto1KimandMarsh2004).AsdescribedinJohnsonetal.(2009),fragmentsweresequentiallymatchedtopatternsasdescribedpreviously(Kentetal.2003;Fig.2a).Acompressionutilitywasthenusedtoremovemultiplematchestothesameorganismandtocombinepatternsthatmatchedmultipleorganisms(Fig.2b).Aphylogeneticsortingalgorithmmatchingthe16SrDNAgenesintheNationalCenterforBiotechnologyInformation(NCBI)sequencedatabasewasthenappliedtothefinalpatternset(Fig.2c).AlthoughtheidentifiedpatternscanbereportedbyInSilico?atfivedifferentphylogeneticlevels(phylum,order,class,family,Fig.2),thedatainthisstudyarepresentedatthegenuslevelbecausethisgivesinformationonthehydrogensulfideproductioncapabilitiesofeachtypeofmicroorganism.The“unclassified”groupofpatternsdiffersfromunmatchedfragmentsbecauselittleornophylogeneticinformationisdepositedwiththeirrespectivesequence.Unclassifiedpatternsmakeupalargepercentageofthedatasincesequencesareoftendepositedwithoutsufficientphylogeneticidentification.Sequencesinthe“multipleclassification”categoryfitintomorethanonedistinctphylogeneticgroupandconsistofalltheuniquespeciesfromphylogeneticallydifferentgroupsthatmatchasinglefragmentpattern.FurtherphylogeneticassignmentGiventherelativelyhighpercentageoffragmentpatternsfromtheInSilico?outputthatwereeitherinthe“Multiples”or“Unclassified”categories,furtheranalysiswasconducted.Anyfragmentpatternthatwasassignedtoeitherthe“Multiples”or“Unclassified”categorywasthenreanalyzedbyenteringitintotheBLASTtoolintheNCBInucleotidedatabase(http://www.ncbi.nlm.nih.gov/nucleotide/),andtheretrievednucleicacidsequencelistedinNCBIwasthenenteredintotheMichiganStateUniversityRibosomalDatabaseProjectclassifierwebsite(http://rdp.cme.msu.edu/classifier/classifier.jsp)usinga95%confidencethresholdforphylogeneticassignment(Johnsonetal.2009;Wangetal.2007).Theclassifiertakesasequenceandassignsittothelowesttaxonomiclevelpossiblewithinacertaindegreeofconfidence.Ifthatorganismwasidentifiedwith95%confidenceorhigher,itwasremovedfromthe“Unclassi-fied”categoryandwasreclassified.Insomecases,allofthefragmentpatternsfroma“multiple”categorizationcouldbeanalyzedusingBLASTApplMicrobiolBiotechnolFig.2FlowdiagramdescribingtheInSilicosoftware(InSilicoLLC,Fuquay-Varina,NC)processusedtogenerateTRFLPcommunityprofiles(Johnsonetal.2009).aFragmentsgeneratedfromseparaterestrictiondigestsaresequentiallymatchedtopatternsfoundintheInSilicodatabase.bMatchedpatternsarecompressedbyremovingextraneouspatternsthatbelongtoasingleorganism,asinorganismA,orbycombiningmultipleorganismsthathavethesamepattern,asinorganismCandorganismD.cPatternsarethenmatchedtophylogeneticinformationintheInSilicosoftwareandreportedatfivedifferentlevels.dInthefinaloutputoftheInSilicosoftwarepackage,organismsareidentifiedbygenera.NotethatthepatternbelongingtoorganismsCandDisidentifiedas“multiple”becausethetwoorganismsbelongtodifferentclasses.OrganismFisdefinedasunclassifiedbecausenophylogeneticinformationisavailableforthepatternandtheRDPClassifierandwereidentifiedasthesameorganism.Whenthisoccurred,andtheorganismwasidentifiedwith95%confidenceorhigher,itwasremovedfromthe“Multiples”categoryandwasreclassified(Wangetal.2007).Somefragmentpatternsthatwereoriginallylabeledas“Multiples”containedonlyorganismsthathavebeenidentifiedasfecalcoliformsandGram-negativeentericpathogens(CommitteeonIndicatorsforWaterbornePathogens2004).Therefore,aseparate“Gram-negativeentericbacteria”categorywascreatedforthose“Multiples,”whichincludedthefollowinggenera:Escherichia,Klebsiella,Proteus,Salmonella,and/orShigella.Statisticalmethodstocompareculture-basedandTRFLPresultsKappatestsofagreementwereperformedcomparingthegenerafoundintheculture-basedmethodtothosefoundintheTRFLPoutput.Kappatestsofagreementareameasureofassociation(correlationorreliability)betweentwomeasurementsofthesameitemwhenthemeasurementsarecategorical.Valuescloserto0indicateslighttolittleagreement,whilevaluescloserto1indicatestrongagreement(LandisandKoch1977).ResultsCulture-basedbiochemicaldetectionAdiversegroupofmicroorganismswasisolatedandculturedfromtheH2S-positivesamplevolumesculturedfromsewagesamples,with24differentspeciesisolated.Citrobacterfreundii,E.coli(averageconcentrationof1.23E+07E.coli/100mL),andEntericGroup60repre-sentedmorethan50%ofthetotalisolatesasidentifiedbyeithertheEnterotubeortheAPI20Esystem.Inaddition,knownentericpathogenssuchasShigellawereidentified.Importantly,anumberofdifferentpossibleH2S-producingentericmicroorganismswerealsoisolated,including:Acinetobacterwolffii,Aeromonashydrophila,C.freundii,Klebsiellaozonae,Klebsiellapneumoniae,Salmonellasp.,andYersiniaenterocolitica.Figure3showstheextentofdetectionofH2S-producingmicroorganismsbytheH2Stest
联系客服：cand57</