TodeterminecancerpathwayactivitiesinninetypesofprimarytumorsandNCI60celllines,weappliedaninsilicoapproachbyexamininggenesignaturesreflectiveofconsequentpathwayactivationusinggeneexpressiondata.SupervisedlearningapproachespredictedthattheRaspathwayisactivein～70%oflungadenocarci-nomasbutinactiveinmostsquamouscellcarcinomas,pulmonarycarcinoids,andsmallcelllungcarcinomas.Incontrast,theTGF-β,TNF-α,Src,Myc,E2F3,andβ-cateninpathwaysareinactiveinlungadenocarcinomas.WepredictedanactiveRas,Myc,Src,and/orE2F3pathwayinsignificantpercentagesofbreastcancer,colorectalcarcinoma,andgliomas.OurresultsalsosuggestthatRasmaybethemostprevailingoncogenicpathway.Additionally,manyNCI60celllinesexhib-itedagenesignatureindicativeofanactiveRas,Myc,and/orSrc,butnotE2F3,β-catenin,TNF-α,orTGF-βpathway.Toourknowledge,thisisthefirstcom-prehensivesurveyofcancerpathwayactivitiesinninemajortumortypesandthemostwidelyusedNCI60celllines.The“geneexpressionpathwaysignatures”wehavedefinedcouldfacilitatetheunderstandingofmolecularmechanismsincan-cerdevelopmentandprovideguidancetotheselectionofappropriatecelllinesforcancerresearchandpharmaceuticalcompoundscreening.
