Dear All: I am using DEXSeq for comparing two groups: control and treatment, and I think the signal in the treatment is strong enough to get many significant results on differential exon usage. However, among the ~15,000 genes (after filtering low counts) considered, only ~100 genes are associated with DEU by a FDR control of 0.1. The number of raw p-values < 0.1 is ~5,800. This is obtained by table ( tapply( res2$pvalue < 0.1, geneIDs(ExonCountSet), any ) ) # FALSE TRUE # 1364 5802 By checking the histogram of raw p-values of exons (NOT genes), I find that it is monotonically increasing from 0 to 1, with relatively few counting bins falling into the bins from 0 to 0.2. My question is: does this result make sense in terms of DE testing? Among 240,000 counting bins, only 14,000 are with p-value < 0.1. If I randomly sample, then there would be ~240,000*0.1 = 24,000 counting bins called DE, so the result is really conservative (I think the signal of treatment is pretty strong so the result is weird). Can I know if there is any explanation for this situation? Thank you so much! Best, Gu -- output of sessionInfo(): R version 3.0.1 (2013-05-16) Platform: x86_64-unknown-linux-gnu (64-bit) locale: [1] LC_CTYPE=en_US LC_NUMERIC=C LC_TIME=en_US [4] LC_COLLATE=en_US LC_MONETARY=en_US LC_MESSAGES=en_US [7] LC_PAPER=C LC_NAME=C LC_ADDRESS=C [10] LC_TELEPHONE=C LC_MEASUREMENT=en_US LC_IDENTIFICATION=C attached base packages: [1] parallel stats graphics grDevices utils datasets methods [8] base other attached packages: [1] DEXSeq_1.6.0 Biobase_2.20.0 BiocGenerics_0.6.0 loaded via a namespace (and not attached): [1] biomaRt_2.16.0 Biostrings_2.28.0 bitops_1.0-5 [4] GenomicRanges_1.12.4 hwriter_1.3 IRanges_1.18.1 [7] RCurl_1.95-4.1 Rsamtools_1.12.3 statmod_1.4.17 [10] stats4_3.0.1 stringr_0.6.2 tools_3.0.1 [13] XML_3.96-1.1 zlibbioc_1.6.0 -- Sent via the guest posting facility at bioconductor.org.

Hi On 23/07/13 14:47, Gu [guest] wrote: > By checking the histogram of raw p-values of exons (NOT genes), I > find that it is monotonically increasing from 0 to 1, with relatively > few counting bins falling into the bins from 0 to 0.2. You are right, DEXSeq sometimes tends to be overly conservative, which then results in a skewed p value histogram as you describe it. Usually, it is, however, only a rather slight skew, and it seems that the performance is unusually bad for your specific dataset. The main reason for the conservative results is the way we estimate dispersion. Since the release of DEXSeq, we have made quite some progress in improving the dispersion estimation by now using an empirical-Bayes shrinkage estimator, and DESeq2 now offers a much better solution, at least for gene-level tests. We are working on applying the same changes to DEXSeq, and this should solve your issue. I'm afraid, however, that I have to ask you for some patience until we are finished with these changes. Simon