I have a dataset with many classes and would like to find the genes associated (differentially expressed) with each class. The experimental design is two-color with a common reference. I am not interested in differential expression between the reference and the sample, but only the differences between samples. One approach would be to do T-tests between ratios for each group and all others. Another would be to use F-statistics to determine those genes that are differentially expressed in a more general sense. What I really want, I think, is a combination of the two so that I can label each gene with class(es) and know that it is differentially expressed across samples. What functions exist for doing something like this? I think the limma package provides ClassifyTestsF which looks useful. Are there others? Have others used other approaches? Thanks, Sean -- Sean Davis, M.D., Ph.D. National Institutes of Health Postdoctoral Research Fellow National Human Genome Research Institute Clinical Fellow National Cancer Institute Johns Hopkins University Clinical Fellow Department of Pediatric Oncology --