Dear BioC community, I've been using a weighted Stouffer method to combing p-values (one- tailed) derived from six hgu133a microarray experiments performed by six different labs. Each experiment roughly address the same null hypothesis: addition of receptor ligand for 24h has no effect on gene expression. I have then taken the combined Stouffer z-score and transformed it back to a two- tailed p-value. #z is a vector of combined z-scores, one for each probe set in the hgu133a array. >combP<-2*(pnorm(abs(z),lower.tail=FALSE)) My question is: Is it valid to apply multiple correction to combP similar to something like: >FDRcombP<-p.adjust(combP,method="BH") If I were doing a Venn overlap of significant probe sets from these six experiments I would be using probe sets which has passed an FDR corrected p-value threshold. My gut is telling me to adjust the Stouffer derived significant score for multiple testing as well. Thanks for any feedback... Scott A. Ochsner, Ph.D. NURSA Bioinformatics Molecular and Cellular Biology Baylor College of Medicine Houston, TX. 77030 phone: 713-798-6227

Dear Scott, As far as I can tell your gut feeling is correct. Obviously the Meta-Analysis type approach you are using makes certain assumptions (the studies are independent, testing the same hypothesis etc...), but assuming those are met the the p-value you obtain from the p-value combination method should be a valid one, i.e it is uniformly distributed under the Nullhypothesis.. I see no reason why you shouldn't use the BH method on this vector of combined p-values (apart from the problems that you have anyway, i.e. dependence between genes, but that is an issue that already arises for a single experiment). The only thing one should NOT do here is first to adjust p-values and then combine the adjusted p-values! that would give you a result that I wouldn't be able to interpret easily. Best Wishes Claus Ochsner, Scott A wrote: > Dear BioC community, > > I've been using a weighted Stouffer method to combing p-values (one- tailed) derived from six hgu133a microarray experiments performed by six different labs. Each experiment roughly address the same null hypothesis: addition of receptor ligand for 24h has no effect on gene expression. I have then > taken the combined Stouffer z-score and transformed it back to a two-tailed p-value. > > #z is a vector of combined z-scores, one for each probe set in the hgu133a array. > >> combP<-2*(pnorm(abs(z),lower.tail=FALSE)) >> > > My question is: Is it valid to apply multiple correction to combP similar to something like: > >> FDRcombP<-p.adjust(combP,method="BH") >> > > If I were doing a Venn overlap of significant probe sets from these six experiments I would be using probe sets which has passed an FDR corrected p-value threshold. My gut is telling me to adjust the Stouffer derived significant score for multiple testing as well. > > Thanks for any feedback... > > > Scott A. Ochsner, Ph.D. > NURSA Bioinformatics > Molecular and Cellular Biology > Baylor College of Medicine > Houston, TX. 77030 > phone: 713-798-6227 > > _______________________________________________ > 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 > > > > > > > Click link below to report this email as spam. > https://www.mailcontrol.com/sr/e4yaOtd+u0gGcUpnapneyFP5samX2xOHgTnL8 0JX1y3!tv1zue6U8TwiGm5I6NGLU1EnlruGv93v6VtG4uN32BtCsu8iMKdTxOsFmzBFyCJ IVRlBJNp9n6EkZ4K3A1fzHEM2Qdc5hYUoMNE1cKVkKg6aTt0Owfxd8CJ1s8FXBm+eUaYEE T4Thyz2KjkCN8omW0enYj13SFUGlvJH6newm+J4u9Ekoi!I > > > -- ********************************************************************** ************* Dr Claus-D. Mayer | http://www.bioss.ac.uk Biomathematics & Statistics Scotland | email: claus at bioss.ac.uk Rowett Research Institute | Telephone: +44 (0) 1224 716652 Aberdeen AB21 9SB, Scotland, UK. | Fax: +44 (0) 1224 715349