Hi, Apologies if this is a naive limma question but the experimental design in this analysis is quite confusing, which has made me want to confirm the approach with the experts! I have the following set of Affy CEL files that are from 19 treatment samples and 5 control samples. My targets file is as follows: file expt --------- ----- Batch1_1.CEL s_1 Batch1_2.CEL s_2 Batch1_3.CEL s_3 Batch1_4.CEL s_4 Batch1_5.CEL s_5 Batch1_6.CEL s_6 Batch1_7.CEL control_T Batch1_9.CEL control_B Batch1_11.CEL control_P Batch2_2.CEL s_7 Batch2_3.CEL s_8 Batch2_5.CEL s_9 Batch2_6.CEL s_10 Batch2_7.CEL s_11 Batch2_9.CEL s_12 Batch2_10.CEL s_13 Batch2_11.CEL s_14 Batch2_12.CEL s_15 Batch2_13.CEL s_6 Batch2_14.CEL control_P Batch2_15.CEL control_B Batch2_16.CEL control_T Batch3_1.CEL control_P Batch3_2.CEL control_T Batch3_4.CEL control_B Batch3_6.CEL control_O Batch3_8.CEL control_O_LP Batch3_9.CEL s_16 Batch3_10.CEL s_17 Batch3_11.CEL s_18 Batch3_12.CEL s_13 Batch3_13.CEL s_13 Batch3_14.CEL s_19 I want to make several different comparisons of treatment vs control combinations so I am using the makeContrast function. The main issue is how to handle the mixture of technical and biological replicates. For example, I wish to compare samples s_1,s_3,s_5,s_6 and s_7 vs control samples and get the top 100 significant probe sets. So after reading in the targets file and normalising the data... design<-model.matrix(~-1+expt) contr.mat<-makeContrasts(TvsN = (s_1+s_3+s_5+s_6+s_7)/5-(control_T+con trol_P+control_B+control_O+control_O_LP)/5,levels=design) fit <- lmFit(normeset, design ) # normeset is the normalised data fit2 <- contrasts.fit( fit, contr.mat ) fit3 <- eBayes( fit2 ) topTable(fit3, n=100, adjust="BH", sort.by="B") does the above correctly take into account the technical replicates? If not what is the best way to handle this? Thanks for any help, Steve ------------------------- Medical Sciences Division Oxford University