Hi Shailesh, By default, GAGE decomposes the group-wise comparison between treated vs control samples into pair-wise comparisons and summarize the individual p-values from multiple pair-wise comparisons into a global p-value. The gage function implements all 4 types of comparison schemes (through argument compare): 'paired', 'unpaired', '1ongroup', 'as.group'. compare='paired' is the default, i.e. ref and samp are of equal length and one-on-one paired by the original experimental design. If you permute the sample labels, your samples are actually not one-on-one paired by design. You should specify compare='unpaired'. This way, you get the same results not matter how you permute your sample labels within each group. Argument use.fold =TRUE means we use fold change as per gene score (the default). The gage function automatically does group-on-group comparison if use.fold =FALSE (i.e. use t-statistics as per gene score). Therefore, the results did not change with your sample label permutation in each group. The following example will show you the usage of different compare values in your case. HTH. Weijun library(gage) data(gse16873) cn=colnames(gse16873) hn=grep('HN',cn, ignore.case =T) dcis=grep('DCIS',cn, ignore.case =T) set.seed(1) hn2=sample(hn) dcis2=sample(dcis) datakegg.gs) #with default compare="paired", you get different results gse16873.kegg.p <- gage(gse16873, gsets = kegg.gs, ref = hn, samp = dcis) gse16873.kegg.p2 <- gage(gse16873, gsets = kegg.gs, ref = hn2, samp = dcis2) #with compare="unpaired", you get the same results gse16873.kegg.p <- gage(gse16873, gsets = kegg.gs, ref = hn, samp = dcis, compare="unpaired") gse16873.kegg.p2 <- gage(gse16873, gsets = kegg.gs, ref = hn2, samp = dcis2, compare="unpaired") ________________________________________ From: Shailesh Tripathi [stripathi01@qub.ac.uk] Sent: Monday, August 22, 2011 2:51 PM To: Luo, Weijun Subject: RE: GAGE package help required Dear Weijun Luo I am using GAGE for my data analysis. There is a one question regarding the GAGE , I am using 'use.fold = TRUE' option for analysis. I am calculating p -values for pathways using different strategies. I am permuting both treatment and control samples of the data independently for each gene . i.e. I am not mixing samples of control and treatment but just permuting their labels independently (for each condition). I am getting different p-values for pathways which should not be different as GAGE is not doing any sample permutation in order to calculate p-value for pathways. I would like know why this is happening, when I keep 'use.fold = FALSE' p-values of pathways remain same without permutation and after permutation, which is the right answer . Thanks and Regards Shailesh Tripathi