Hello, ? I havea an experiment with mice done on Affymetrix and it looks like this: ? FileName?? Treatment???Batch???MouseID =============================== file1??????????? control??????? 1?????????? 101 file2??????????? control??????? 1?????????? 102 file3??????????? control??????? 1?????????? 103 file4????????????treatA ??????? 1?????????? 104 file5????????????treatA? ?????? 1?????????? 105 file6????????????treatA?? ????? 1?????????? 106 file7??????????? control????????2?????????? 107 file8??????????? control????????2?????????? 108 file9??????????? control????????2?????????? 109 file10???????????treatA????????2?????????? 110 file11???????????treatA????????2?????????? 111 file12???????????treatA????????2?????????? 112 file13??????????? control????????3?????????? 113 file14??????????? control????????3?????????? 114 file15??????????? control????????3?????????? 115 file16 ???????????treatB? ???????3?????????? 116 file17? ??????????treatB?? ??????3?????????? 117 file18?? ?????????treatB??? ?????3?????????? 118 file19??????????? control????????4?????????? 119 file20??????????? control????????4?????????? 120 file21??????????? control????????4?????????? 121 file22????????????treatB?????????4?????????? 122 file23????????????treatB?????????4?????????? 123 file24????????????treatB?????????4?????????? 124 ? The controls from all batches are (should be) biologically the same and I am looking for differentially expressed genes in: a)?treatB versus treatA, b) treatA versus control, c) treatB versus control. ? Are?the design matrix and contrast matrix as follows? ? factor_treatment = factor(Treatment, levels=C("control","treatA","treatB")) factor_batch = factor(Batch, levels=C("1","2","3","4")) design = model.matrix(~0 +?factor_batch + factor_treatment)? fit = lmFit(eset, design) cont.matrix = makeContrast( ????????????????? treatBvsA = (treatB - treatA), ????????????????? treatAvsControl = (treatA - control), ????????????????? treatBvsControl = (treatB - control), ????????????????? levels = design) fit2 = lmFit(fit, cont.matrix) fit2?= eBayes(fit2) ? Cheers, Daniel Berger