Hello, I've compared results from a jusrRMA and a gcrma analysis of several affy mouse chips. Ten chips come from 3 different experimentators (i.e. 3 or 4 replicates) that have worked with the same lab protocol to measure gene expression (same untreated mouse cell lines). I've run justRMA, gcrma (version 0.05 since I'm still running R-1.8.1), and MAS5 in combination with quantile or vsn normalisation. An anova via "Intensity ~ batch" (where batch has 3 levels, one for each experimentator) should ideally give zero significant genes, if they've done a reproducalbe work and the cells behaved well ;-) ... . >From previous analysis I already know that the 3 batches are very different. However, gcrma with mle comes closest to zero (1601 of the 12488 probe sets with p <= 0.01). I'm not realy sure what gcrma actually does. How is the sequence information taken into account? If gcrma is able to minimize the differences between cross-laboratory differneces in my experiment, this would mean the sequence locations themselfs account for the differences, which is strange, since they're the same or the chips. I'd be happy if someone could give me a hint how gcrma works? thanks a lot for your comments, Arne -- Arne Muller, Ph.D. Toxicogenomics, Aventis Pharma arne dot muller domain=aventis com