Hidly-ho bioconductorinos, This question is more a matter of methodology rather than how to use bioconductor. If there is a better mailing list for this then I'd be happy to resubmit this there. I've looked for an answer to this question in the bioconductor website; however, it doesn't seem that anyone has asked this particular question just yet. I'm analyzing an Agilent microarray experiment for someone. The purpose of the experiment is to identify how a knockout (ko) is differently affected than the wild type by a treatment (treatment=t, untreated=u). Therefore, the targets file is as follows: array Cy3 Cy5 1 wt.t wt.u 2 wt.u wt.t 3 wt.t wt.u 4 ko.u ko.t 5 ko.t ko.u 6 ko.u ko.t As you can see, there were no hybridizations between wt and ko, so not all of the effects can be measured. So what I've decided to do instead was simply to perform a t-test between the M values from the wt arrays and the M values from the ko arrays (or another - likely better - hypothesis testing approach that someone may suggest). My question is about the dye swap (see arrays 2 and 5). Usually, limma identifies the dye swap when the design matrix is passed to lmFit. Since I won't be using lmFit, I'm wondering how I can handle the dye swap. Do I simply take the opposite sign of the M values from arrays 2 and 5, and if so should I do that before normalization? Is there a better method for handling the dye swap in this situation? Any help you can offer would be greatly appreciated! Thanks, Wyatt K. Wyatt McMahon, Ph.D. Postdoctoral Research Associate Department of Internal Medicine 3601 4th St. - Lubbock, TX - 79430 P: 806-743-4072 F: 806-743-3148 [[alternative HTML version deleted]]