Dear Bioconductor users, Firstly I'm sorry if this subject is appropriate for the list. I remember there were some discussion in this list on what to do when treatment produces a global shift on gene expression. Now I'm facing this problem on T cell activation experiment. Upon activation T cells change dramaticaly. They start proliferation, getting much bigger and make more RNA. After running affy chips, whatever normalizations I did, thousands genes(20%~30%) on chip changed. Of course they all give equal number of up- and down-regulated genes. I feel really hard to make explantation from these data. And I think it's dangerous to normalize using housekeeping genes because HKs also change after T cell activation. Could anybody please give me some clues/publications on which pre-process/normalization works better for this situation? -- Dapeng Cui Department of Medicine Emory University School of Medicine Atlanta, GA 30322

This depends on how you call your genes to be affected. e.g. fold change, t-test and at what threshold. What preprocessing did you try ? Alternatively you simply rank the genes and see if the top/bottom 10, 20, 100 etc makes sense to you. AFAIK, Affymetrix preprocessing algorithms do not necessarily have to give equal numbers of up and down regulated genes. In fact in one of the projects that we are involved in, of the final list of 200-300 genes we found 95% of the genes were down regulated. Some of these were successfully verified with taqman and the biologists were happy. I am not a biologist but from what I understand normalization based on house keeping genes are generally not recommended because they may not be house keeping genes in the tissue that you are testing for. Regards, Adai On Fri, 2005-04-08 at 13:10 -0400, Dapeng Cui wrote: > Dear Bioconductor users, > > Firstly I'm sorry if this subject is appropriate for the list. I remember there > were some discussion in this list on what to do when treatment produces a > global shift on gene expression. Now I'm facing this problem on T cell > activation experiment. Upon activation T cells change dramaticaly. They start > proliferation, getting much bigger and make more RNA. After running affy chips, > whatever normalizations I did, thousands genes(20%~30%) on chip changed. Of > course they all give equal number of up- and down-regulated genes. I feel > really hard to make explantation from these data. And I think it's dangerous to > normalize using housekeeping genes because HKs also change after T cell > activation. Could anybody please give me some clues/publications on which > pre-process/normalization works better for this situation? > >