Dear Hong, You can fit any multiple regression model in limma, so the issue isn't one of software but rather of deciding what model you want to fit. And we can't tell you that. One possibility to consider is to fit a regression spline along the time course, rather than treating each time point independently, something like: library(splines) design <- model.matrix(~Drug*ns(Time,df=3)) Best wishes Gordon > Date: Mon, 6 Jun 2011 22:02:16 -0700 (PDT) > From: Hong Nie <hongnie at="" yahoo.com=""> > To: "bioconductor at r-project.org" <bioconductor at="" r-project.org=""> > Subject: [BioC] limma: multiple regression > > Hello all,? > > I am working on a Yeast2 Affymetrix project.? > > The yeast received one drug treatment at 3 concentration (0, 0.5, 1) > with 10 time points for each treatment, without replication in each > combination.? > > eg: > > drug ?time > 0 ? ? ? ? ? 0 > 0 ? ? ? ? ? 6 > 0 ? ? ? ? ? 12 > ... > 0.5 ? ? ? ? 0 > 0.5 ? ? ? ? 6 > 0.5 ? ? ? ?12 > > ...? > 1 ? ? ? ? ? 90 > > My target is to test what genes ?significantly change for the drug > treatment and at which time points.? > > I am using limma package. > > > My question is how I should make the model.matrix.? > > It seems that I can't use ?~ Drug*Time, that will result in no residual > degrees of freedom.? > > Under such condition, may I put Variable Time into the model? Or I have > to do one linear regression (~Drug) for each time point, respectively.? > > Or any other package will work this analysis better? > > Thank you for any suggestion! ?I am a newcomer for Bioconductor.? > > -Hong