Hello I have been involved recently in analyzing some microarray experiments performed with affymetrix chips. This task had been previously done by another analyst who left me some scripts, but no explanations. The procedure she used to follow was first to normalize the arrays, say using rma and then, before doing any tests she used to apply two filters - She kept only those genes whose signal was greater than a threshold on all arrays (she used "log(100)" as this threshold) - Assuming for simplicity that there were only two groups she applied a second filter keeping only those genes where the base-2 logarithm of the difference between the mean of the two groups was greater than 1.5 I think I understand the rationale under this procedure, but also I find it somewhat arbitrary. Could someone please orient me about if this a usual/right way to proceed, or address to some references or examples which help to diminish the feeling of arbitrarity? Thanks for your help ======================== Teresa Casals

Hi, Teresa Casals wrote: > Hello > > I have been involved recently in analyzing some > microarray experiments performed with affymetrix > chips. > > This task had been previously done by another analyst > who left me some scripts, but no explanations. > > The procedure she used to follow was first to > normalize the arrays, say using rma and then, before > doing any tests she used to apply two filters > > - She kept only those genes whose signal was greater > than a threshold on all arrays (she used "log(100)" as > this threshold) > - Assuming for simplicity that there were only two > groups she applied a second filter keeping only those > genes where the base-2 logarithm of the difference > between the mean of the two groups was greater than > 1.5 > > I think I understand the rationale under this > procedure, but also I find it somewhat arbitrary. > Pretty much all filtering of genes is arbitrary. I don't think that there is a way out of that, unless you know a lot about the underlying biology. Some reduction of the genes that were assayed is necessary so you must choose some method. We have found that it is better to filter on variability rather than level (although at one time I was a fan of filtering on level). Choose some (arbitrary amount of variability) and filter out those genes which do not show that amount of variability across samples. You can see the second paper at http://www.bepress.com/bioconductor/ for some more detailed discussions of the issues. Robert > Could someone please orient me about if this a > usual/right way to proceed, or address to some > references or examples which help to diminish the > feeling of arbitrarity? > > Thanks for your help > > ======================== > Teresa Casals > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > -- Robert Gentleman, PhD Program in Computational Biology Division of Public Health Sciences Fred Hutchinson Cancer Research Center 1100 Fairview Ave. N, M2-B876 PO Box 19024 Seattle, Washington 98109-1024 206-667-7700 rgentlem at fhcrc.org