I am hoping for some advice with GeneMeta: I am combining a two color and single color microarray. I'm a little confused by the way that GeneMeta handles two color data? Specifically getdF : what if the expression set object is from a two color array , say for example a common reference design, how should the class vector indicating the group membership be constructed? For example a 2 color experiment; > class(MA.t) [1] "MAList" attr(,"package") [1] "limma" library(convert) MA.exp.t<-as(MA.t,"ExpressionSet") ### get expression set object for GeneMeta pData(MA.exp.t) > pData(MA.exp.t) FileName Data from 070503 Data from 070503.txt Data from 08AND16 Data from 08AND16.txt Data from 260603A Data from 260603A.txt Data from 260603B Data from 260603B.txt Data from 260603C Data from 260603C.txt Data from 260603D Data from 260603D.txt Data from 260603E Data from 260603E.txt Data from CONvsBAY2 Data from CONvsBAY2.txt > targets FileName Name Cy3 Cy5 70503 Data from 070503.txt G Ref BA 08AND16 Data from 08AND16.txt H Ref BA A Data from 260603A.txt A Ref BA B Data from 260603B.txt B Ref BA C Data from 260603C.txt C Ref BA D Data from 260603D.txt D Ref BA E Data from 260603E.txt E Ref BA CONvsBAY2 Data from CONvsBA2.txt F Ref BA > dim(MA.exp.t) Features Samples 11629 8 All arrays are of the same class the ratio of affected vs control so for example: Split2<-MA.exp.t Split2.class<-c(1,1,1,1,1,1,1,1) d.Split2 <- getdF(Split2, Split2.class) ## this fails using Split2.class<-c(1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0) ## fails as dim(eset)[2] != length(Split2.class) is tripped in the function call.... ______________________________________________________________________ __________________ So I have replaced getdF with a direction calculation of the t-statistic (since I had it anyway: Split2.t.stat2<-fit2.agilent2.t$coef[,1] / fit2.agilent2.t $stdev.unscaled[,1] / fit2.agilent2.t$sigma ## fit2.agilent is unmoderated t-stat.. al la limma the gives the same results as getdF for "typical" single color arrays) then proceed as follows.... which I think is correct... Split2.class<-c(rep("BA",times=dim(Split2)[2]),rep("Ref",times=dim(Spl it2)[2])) Split2.class<-factor(Split2.class) levels(Split2.class) <- c(0,1) Split2.class.counts<-tapply(Split2.class,Split2.class,length) d.Split2<-Split2.t.stat*sqrt(sum(Split2.class.counts)/prod(Split2.clas s.counts)) ## as per authors paper d=t.stat * sqrt(n1+n2/n1*n2) BUT getdF is also used the the function zScores and zScoresFDR . Both these only require minor modifications to replace but I'm a concerned that I have missed something , can two-color data can be handled without modification? Thanks Paul [[alternative HTML version deleted]]