The question of whether normalizing by condition is reasonable frequently comes up on this list. For a recent talk, I took the first 2 conditions (experiments) of the Affymetrix Latin Square spikein experiment and normalized by RMA all together and within treatment. The experiment uses identical RNA except for 42 spike-ins which are varied by condition. There are 3 replicates, making 6 arrays in all. The number of "significant" genes at p<.01 using limma with eBayes (but no multiple comparison adjustment) is 144 (capturing most but not all of the spike-ins) when the arrays are normalized together and 14374 when they are normalized by experiment. Naomi S. Altman 814-865-3791 (voice) Associate Professor Dept. of Statistics 814-863-7114 (fax) Penn State University 814-865-1348 (Statistics) University Park, PA 16802-2111

Hi Naomi I would tend to agree with you. However, there are numerous (ok, some) applications where the (just assume simple treatment/control setup) treatment is expect to be massively different than the control. Say eg. if treatment is cyclohexamid. As basically all normalization techniques were developed for the "largely all genes unchanged" situation, it is not clear how one would approach a massive global change. However, come to think of it, the examples I have in mind are not exactly standard expression studies. In fact they are not even expression studies. /Kasper On Feb 27, 2006, at 7:30 PM, Naomi Altman wrote: > The question of whether normalizing by condition is reasonable > frequently comes up on this list. For a recent talk, I took the > first 2 conditions (experiments) of the Affymetrix Latin Square > spikein experiment and normalized by RMA all together and within > treatment. The experiment uses identical RNA except for 42 spike- ins > which are varied by condition. There are 3 replicates, making 6 > arrays in all. > > The number of "significant" genes at p<.01 using limma with eBayes > (but no multiple comparison adjustment) is 144 (capturing most but > not all of the spike-ins) when the arrays are normalized together and > 14374 when they are normalized by experiment. > > > Naomi S. Altman 814-865-3791 (voice) > Associate Professor > Dept. of Statistics 814-863-7114 (fax) > Penn State University 814-865-1348 > (Statistics) > University Park, PA 16802-2111 > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor