Dear list, I have got 2x3 arrays (3 wildtype, 3 mutant) of the type MOE 430 2.0 (mouse genome). Since there are so many different approaches to normalization, I do not really know which one to choose. I thought about using GCRMA, since it is said to be good for low expressed genes and I have quite a lot of those (double knockout experiment). Also, I thought GCRMA would be better than, say, VSN for my design, since parameter estimation out of 3 chips doesn't seem to be very trustworthy. I got a few questions concerning GCRMA, though. When looking at the histogram of the variances (across all arrays), I see a clear bimodal distribution. After searching in the mail archives here, I found some hints at GCRMA being only optimised for a certain human chip. Is that true? Should I rather be using another method? Any suggestions on this? In general, the quality control plots of the gcrma normalized values seem okay to me. I looked at variance to mean expression value, wildtype versus mutant variances scatterplot, boxplot and others. When comparing the histograms of the expression values or variances to other normalization methods (rma, vsn, mas5), I get obviously different distributions. Is that good or bad? Any suggestions on how to decide on this? Thanks a lot for helping out. Best, ww ______________________________________________________________________ ___ Mit der Gruppen-SMS von WEB.DE FreeMail k?nnen Sie eine SMS an alle