Hi all, sorry if I write such a "not strictly technical" question to this mailing list. Even if not linked to the programming, maybe somebody would be so nice to comment on my issue. On a dataset I performed a simple DE analysis between two groups through limma (eBayes and topTable). The histogram of the p-values are quite nice, showing a good number DE genes: around 400 with adj p-vals ranging from 0.01 to 0.05. Not extremely highly sign. but still.... The highest FC is around 1.5. I also ran SAM and I got almost the same results. Am I correct interpreting in a general way like there many "small" significative differences? On the same data I also ran a Predictive Analysis through PAM. By minimizing the CV errors I can select around 70 genes, ALL of them are also present at the top positions into the list obtained with the DE analysis. The only thing that while I don't find it surprising but still puzzles me, is the high level of the errors. The confusion matrix gives me a level of misclassification error of 0.1 and 0.6 for the two classes, and an overall error of 0.3 Please, can somebody help me comment on this? Do the error levels of the PAM analysis invalidate in some way the results obtained with the two DE methods? Or simply due to the small quantitative differences the Predictive analysis is not able to clearly classify the selected genes? Or maybe something else? Thanks a lot in advance, any help will be appreciated, Giulio. [[alternative HTML version deleted]]

Mhhh, it seems that maybe I wasn't clear, or my previous it passed unnoticed. Please allow me to repost it (see below this email). In short: Quite clearly small differentiate genes with Differential Expression. Same results, same genes with PAM, but high CV errors! Is it PAM """failing""", by overestimating the errors (because of the small, even if significative, differences)? Any clues about why I get on one side high misclassification errors for the same genes that on the other side are nicely and clearly differentially expressed? Sorry for the repost, it is the first and last time, Best Giulio > From: perimessaggini@hotmail.com > To: bioconductor@stat.math.ethz.ch > Date: Fri, 24 Sep 2010 14:55:39 +0000 > Subject: [BioC] Puzzled about limma and PAM results validity > > > > Hi all, sorry if I write such a "not strictly technical" question to this mailing list. Even if not linked to the programming, maybe somebody would be so nice to comment on my issue. > > On a dataset I performed a simple DE analysis between two groups through limma (eBayes and topTable). The histogram of the p-values are quite nice, showing a good number DE genes: around 400 with adj p-vals ranging from 0.01 to 0.05. Not extremely highly sign. but still.... The highest FC is around 1.5. I also ran SAM and I got almost the same results. > Am I correct interpreting in a general way like there many "small" significative differences? > > On the same data I also ran a Predictive Analysis through PAM. By minimizing the CV errors I can select around 70 genes, ALL of them are also present at the top positions into the list obtained with the DE analysis. > The only thing that while I don't find it surprising but still puzzles me, is the high level of the errors. The confusion matrix gives me a level of misclassification error of 0.1 and 0.6 for the two classes, and an overall error of 0.3 > > Please, can somebody help me comment on this? Do the error levels of the PAM analysis invalidate in some way the results obtained with the two DE methods? Or simply due to the small quantitative differences the Predictive analysis is not able to clearly classify the selected genes? Or maybe something else? > > Thanks a lot in advance, > > any help will be appreciated, > > Giulio. > > > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor [[alternative HTML version deleted]]