Hi list, I thought I would like to get some feedback from you guys with regards to analysis of data from Affymetrix Human Exon St1.0 array The research question I have in mind is this: In a list of 200 candidate genes associated with a disease, how many of these are alternatively spliced. I have a total of 30 arrays (21 cases and 9 controls), and I want to study the candidates to see if there are alternative splicing events in each of the gene, and where at. This is my current analysis method: 1. RMA normalize the CEL files- I do this in APT using the FULL annotations from Affymetrix. This was done in APT simply because doing it in R took really long (I don't have access to a server), and APT allows me to also normalize at the transcript level. 2. Once the normalization is done in APT, I will get a text file with all the probeset summaries. This text file is then imported into R, along with the Affymetrix annotations. 3. I then isolate the probes of the gene/genes of interest. I used the Bioconductor package Biomart (thanks for this great package, btw!) to isolate the probes of interest using the %in% command in R. 4. Concurrently, I also isolate the transcript level normalized scores for the probesets in the gene/genes. This is done by looking up the transcript cluster ID in the annotation based on the probes of interest that was isolated in (3). 5. The probeset level scores are then normalized for gene expression by taking the log2 ratio of each scores, so that the normalized score is log2(probeset) - log2(transcript cluster). I do this because I want to account for different probeset intensity which may have arose because of different gene level expression. 6. To then determine if the cases and control differ, I then do ANOVA at the probeset level to see if there is significant difference between the case and the control. I was wondering, though, if this is the best way to test for alternative splicing? Or perhaps, you might want to point out some other equally viable, and perhaps more rigorous, method for me to do this? Thanks for your input! Jeremy [[alternative HTML version deleted]]