简介：Domaindatabaseisessentialfordomainpropertyresearch.Eliminatingredundantinformationindatabasequeryisveryimportantfordatabasequality.Herewereportthemanualconstructionofanon-redundanthumanSH2domaindatabase.Thereare119humanSH2domainsin110SH2-containingproteins.HumanSH2swerealignedwithClustalX,andahomologoustreewasgenerated.Inthistree,proteinswithsimilarknownfunctionwereclassifiedintothesamegroup.Someproteinsinthesamegrouphavebeenreportedtohavesimilarbindingmotifsexperimentally.Thetreemightprovidecluesaboutpossiblefunctionsofhypotheticalproteinsforfurtherexperimentalverification.

简介：Microarrayhasbecomeapopularbiotechnologyinbiologicalandmedicalresearch.However,systematicandstochasticvariabilitiesinmicroarraydataareexpectedandunavoidable,resultingintheproblemthattherawmeasurementshaveinherent"noise"withinmicroarrayexperiments.Currently,logarithmicratiosareusuallyanalyzedbyvariousclusteringmethodsdirectly,whichmayintroducebiasinterpretationinidentifyinggroupsofgenesorsamples.Inthispaper,astatisticalmethodbasedonmixedmodelapproacheswasproposedformicroarraydataclusteranalysis.TheunderlyingrationaleofthismethodistopartitiontheobservedtotalgeneexpressionlevelintovariousvariationscausedbydifferentfactorsusinganANOVAmodel,andtopredictthedifferentialeffectsofGV(genebyvariety)interactionusingtheadjustedunbiasedprediction(AUP)method.ThepredictedGVinteractioneffectscanthenbeusedastheinputsofclusteranalysis.Weillustratedtheapplicationofourmethodwithageneexpressiondatasetandelucidatedtheutilityofourapproachusinganexternalvalidation.