Hello dear List, I conducted an analysis with EdgeR and I thank you for this software. I have a list of DE genes after treatment A and one after treatment B. using the cntrast fucntion, I also have the comparison of A and B, which shows 43 differentially regulated genes. When I check FDR and logFC fot those 43 genes, they are signicant. But when I come back to list A or B, I see that some of those 43 genes did not reach significance either for FDR or for logFC, in either list A or B. Should I then get rid of those genes, or FDR in the contrast output takes that into account ? thanks a lot for you answers, best regards, anna -- output of sessionInfo(): R version 2.15.1 (2012-06-22) Platform: i386-pc-mingw32/i386 (32-bit) locale: [1] LC_COLLATE=French_France.1252 LC_CTYPE=French_France.1252 [3] LC_MONETARY=French_France.1252 LC_NUMERIC=C [5] LC_TIME=French_France.1252 attached base packages: [1] stats graphics grDevices utils datasets methods base -- Sent via the guest posting facility at bioconductor.org.

hi again everybody, actually log FC is not a worry because I can choose a cut-off myself ( logFC>1 for example), the worry is more about FDR that are >0.05 in list A or B, but still the gene has FDR<0.05 in the contrast list, so is that gene significantly differently regulated ? thanks, anna -- output of sessionInfo(): R version 2.15.1 (2012-06-22) Platform: i386-pc-mingw32/i386 (32-bit) locale: [1] LC_COLLATE=French_France.1252 LC_CTYPE=French_France.1252 [3] LC_MONETARY=French_France.1252 LC_NUMERIC=C [5] LC_TIME=French_France.1252 attached base packages: [1] stats graphics grDevices utils datasets methods base -- Sent via the guest posting facility at bioconductor.org.

Anna, I found your question difficult to interpret. If I understand your question correctly, you have compared a treatment A back to a control treatment, and you have also compared treatment B back to the control. So you have contrasts AvsControl and BvsControl. Now you make the contrast BvsControl-AvsControl, and you find some genes significant that were not significant for either BvsControl or AvsControl. This is not a mistake, and it's no great mystery. It can easily be that B is higher than the Control for some gene, but not by enough to reach signficance, and A is less the control for the same gene, but not by enough to reach significance, but the combined difference B vs A is large enough to be significantly greater than zero. This is just because B-A is larger than either B-Control or A-Control. In such cases, you know that A and B are different, but it is not necessarily discernible whether both or just one are different from the control. How you interpret such genes depends on your scientific questions. We can't tell you how to do that. Best wishes Gordon PS. Please check over the posting guide, which has been expanded recently: http://www.bioconductor.org/help/mailing-list/posting-guide/ --------- original message -------- [BioC] EdgeR: about FDR anna [guest] guest at bioconductor.org Sat Nov 3 14:22:38 CET 2012 Hello dear List, I conducted an analysis with EdgeR and I thank you for this software. I have a list of DE genes after treatment A and one after treatment B. using the cntrast fucntion, I also have the comparison of A and B, which shows 43 differentially regulated genes. When I check FDR and logFC fot those 43 genes, they are signicant. But when I come back to list A or B, I see that some of those 43 genes did not reach significance either for FDR or for logFC, in either list A or B. Should I then get rid of those genes, or FDR in the contrast output takes that into account ? thanks a lot for you answers, best regards, anna -- output of sessionInfo(): R version 2.15.1 (2012-06-22) Platform: i386-pc-mingw32/i386 (32-bit) locale: [1] LC_COLLATE=French_France.1252 LC_CTYPE=French_France.1252 [3] LC_MONETARY=French_France.1252 LC_NUMERIC=C [5] LC_TIME=French_France.1252 attached base packages: [1] stats graphics grDevices utils datasets methods base ______________________________________________________________________ The information in this email is confidential and intend...{{dropped:4}}

Hello James, (Hello Gordon) thank you very much for your help,(both of you). I understand pretty well what you mean James,thank you. I was just wondering how we can build certainty on 2 uncertainties. But I check back my data , actually there was not absence of significance at the same time for both treatments, it only happened that one of them at a time was not significantly regulating some genes(either same or opposite direction as the other treatment). So then if I assume that for those genes in that treatment logFc=0, I still had a large enough difference between treatment A and B, which reaches significance. Have a nice day James, thanks again, anna -- output of sessionInfo(): R version 2.15.1 (2012-06-22) Platform: i386-pc-mingw32/i386 (32-bit) locale: [1] LC_COLLATE=French_France.1252 LC_CTYPE=French_France.1252 [3] LC_MONETARY=French_France.1252 LC_NUMERIC=C [5] LC_TIME=French_France.1252 attached base packages: [1] stats graphics grDevices utils datasets methods base -- Sent via the guest posting facility at bioconductor.org.