Dear list, I know there is a gseattperm() function available in the Category package used to perform GSEA test on two group of samples. The permutation is based on phenotype labels of the samples. I am wondering is there a similar function can be used to do the same test, but based on the permutation of genes because I have a very small number of samples ( 6 samples in total for two phenotypes ( 4 versus 2) ). If not, may I get around this problem by transposing my data set to use the gseattper() by permuting on genes? Thank you very much in advance! Kind regards, Yuan

Hi Yuan, Yuan Hao wrote: > Dear list, > > I know there is a gseattperm() function available in the Category > package used to perform GSEA test on two group of samples. The > permutation is based on phenotype labels of the samples. I am wondering > is there a similar function can be used to do the same test, but based > on the permutation of genes because I have a very small number of > samples ( 6 samples in total for two phenotypes ( 4 versus 2) ). If not, > may I get around this problem by transposing my data set to use the > gseattper() by permuting on genes? Thank you very much in advance! Not really. If you transpose, then you will be computing t-tests between genes within samples rather than between samples within genes. In other words, the function won't know that you have transposed the data, so won't know to permute genes but still compute the t-statistics on a by-gene basis. Anyway, I am not sure this is a reasonable thing to do. Under the null distribution you can argue that the samples are exchangeable because under the null there aren't any differences between samples. However, under the null distribution of no difference between samples, there are still expected to be differences between genes, so they are not exchangeable. Therefore, the distribution you would create by permuting genes would not necessarily correspond to the expectation of no differences between samples. You are probably better off using the t-distribution as your null rather than permuting genes. One way you could do this is to use the 'J-G' statistic proposed in Extensions to gene set enrichment. Jiang Z, Gentleman R. Bioinformatics. 2007 Feb 1;23(3):306-13. Epub 2006 Nov 24. and explained in a more accessible manner in Gene set enrichment analysis using linear models and diagnostics. Oron AP, Jiang Z, Gentleman R Bioinformatics. 2008 Nov 15;24(22):2586-91. Epub 2008 Sep 11. Best, Jim > > Kind regards, > Yuan > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor -- James W. MacDonald, M.S. Biostatistician Douglas Lab University of Michigan Department of Human Genetics 5912 Buhl 1241 E. Catherine St. Ann Arbor MI 48109-5618 734-615-7826