54Geneticcorrelationnetworkpredictionoffores

TheISMEJournal(2018)12:2492–2505●KankanZhao1,2●XiaofeiLv1,2●WeiqinSu1,2●ZhongminDai1,2●JackA.Gilbert3,4●PhilipC.Brookes1,2●KarolineFaust5●JianmingXu1,2Received:9April2018/Revised:11June2018/Accepted:15June2018/Publishedonline:25July2018©TheAuthor(s)2018.ThisarticleispublishedwithopenaccessAbstractSoilecologicalfunctionsarelargelydeterminedbytheactivitiesofsoilmicroorganisms,which,inturn,areregulatedbyrelevantinteractionsbetweengenesandtheircorrespondingpathways.Therefore,thegeneticnetworkcantheoreticallyelucidatethefunctionalorganizationthatsupportscomplexmicrobialcommunityfunctions,althoughthishasnotbeenpreviouslyattempted.Wegeneratedageneticcorrelationnetworkbasedon5421genesderivedfrommetagenomesofforestsoils,identifying7191positiveand123negativecorrelationrelationships.Thisnetworkconsistedof27clustersenrichedwithsetsofgeneswithinspecificfunctions,representedwithcorrespondingclusterhubs.Theclustersrevealedahierarchicalarchitecture,reflectingthefunctionalorganizationinthesoilmetagenomes.Positivecorrelationsmappedfunctionalassociations,whereasnegativecorrelationsoftenmappedregulatoryprocesses.Thepotentialfunctionsofuncharacterizedgeneswerepredictedbasedonthefunctionsoflocatedclusters.Theglobalgeneticcorrelationnetworkhighlightsthefunctionalorganizationinsoilmetagenomesandprovidesaresourceforpredictinggenefunctions.Weanticipatethatthegeneticcorrelationnetworkmaybeexploitedtocomprehensivelydeciphersoilmicrobialcommunityfunctions.IntroductionMicroorganismsoperateattheheartofbiologicalchar-acteristics,biogeochemicalprocesses,andecologyofsoils[1].However,elucidatingthemicrobialfunctionsthatunderpinthesepropertiesofsoilscanbechallenging,primarilyduetothenumericalabundanceofmicrobes[2]andtheirvasttaxonomicandfunctionaldiversity[3]inthesoilenvironment,whichcontainsextremespatialheterogeneityandcomplexchemicalandbiologicalprop-erties[4].Agramofsoilcontainsanaverageof109pro-karyoticcells,and~105distinctprokaryoticgenomes[5].Weestimatethatthemajorityofsoilmicrobialgenomeshaveyettobesequenced[6],assuchwehavelimitedunderstandingofthelinkbetweensoilmicrobialcommu-nitycompositionandfunctionality,whichiscomplicatedbythehorizontalgenetransferpromiscuityofsomebac-teriallineages[7].However,metagenomicsequencingcanprovideasnapshotoftherelativeabundanceofgenesandgenotypes,providinganopportunitytoglimpsesoilmicrobialfunctionalpotential[8–11].Ecosystemsareformedbythehierarchicalorganizationfrompopulations,individuals,pathways,andgenestocommunities[12].Allthemacroscopicpropertiessuchascommunityfunctionsaredependedonhowthemicroscopicbuildingblocks(genes,genotypes,andcells)areassembledandinteract[13].Geneticinteractionsatthecellularscalehavelongbeeninvestigatedinmodelorganisms,especiallyinyeast,foridentifyingfunctionalrelationshipsbetween*[e-mail]1InstituteofSoilandWaterResourcesandEnvironmentalScience,CollegeofEnvironmentalandResourceSciences,ZhejiangUniversity,Hangzhou310058,China2ZhejiangProvincialKeyLaboratoryofAgriculturalResourcesandEnvironment,Hangzhou310058,China3TheMicrobiomeCenter,DepartmentofSurgery,UniversityofChicago,Chicago,IL60637,USA4BioscienceDivision,ArgonneNationalLaboratory,Lemont,IL60439,USA5DepartmentofMicrobiologyandImmunology,RegaInstitute,KULeuven,CampusGasthuisberg,Leuven,BelgiumElectronicsupplementarymaterialTheonlineversionofthisarticle()containssupplementarymaterial,whichisavailabletoauthorizedusers.1234567890();,:1234567890();,:genes[14–16],wherebytheirinteractionsimplythattwogenesshareafunctionalrelationship[17].Studiesexploringtheseinteractionshaveidentifiedmanybiologicalfunctions,includingfunctionaldependencyandredundancy,whicharegovernedbytheinteractionsbetweenseveralenzymes,[18].Althoughonly~1000coregenesintheyeastgenomearelethalwhenmutated,therearearound550,000syntheticlethalgeneticinteractionpairs,includinganextremesetof~10,000geneticinteractionsbetweennon-coregenes[16].However,thecomplexityofgeneticinteractionsinanassemblageatthecommunityscalehasnotyetbeenevaluated.Microbialgeneticsrepresentsauniqueplatformtodeterminewhethergeneticinteractionmodelingcanbeusedtoelucidaterelevantinteractionsbetweengenes.Con-servationofcentralmetabolicfunctionalgenes[19]andhighredundancyoffunctionalgenesinmicrobialcommu-nities[20]suggestthatgenomesofdifferenttaxamightcontainsimilarfunctionalgenes.Therefore,metagenomicdatasetscandisplaystablegeneticinteractionpatternsbecausesimilarfunctionalgenesfromdifferenttaxamighthaveasimilarresponsetoenvironmentalvariations.Withmetagenomics,geneticinteractionscannotbequantifiedusingthecombinatorialconstructionofmutantsbutcanbeexploredusingthepairwisecorrelationcoefficient[21,22].Amatrixofgene-genepairwisecorrelationscanthereforebeusedtosystematicallypredictpotentialgeneticinterac-tionsamonggeneticelementsinadatabase.Networkanalysishasfacilitatedmanydiscoveriesinbothsystemsbiologyandmicrobialeco