Entering edit mode
8.7 years ago
Anushka
▴
20
Is it advisable/(does it make sense) to use different software/R packages (e.g. edgeR, DESeq2, DEGseq, BaySeq etc) for identifying deferentially expressed genes, in order to narrow down the numbers of DEGs (for the better selection of candidate genes), by comparing and finding their intersection (commonly found by all package). The purpose here is NOT to compare different packages, but to check common findings from all.
It's absolutely fine. You can also read recent papers published comparing the different methods to get an idea which one would perform better given a experimental setup.
A problem with such an approach is that one loses any semblance of the statistical certainty of the genes called differentially-expressed. In general, I don't recommend this approach to others or apply it for anything other than instructional purposes myself.
Thanks much @Goutham
As Sean mentioned Its true that you would miss important genes detected by one method and not detected by other. So reading these papers would give an idea which tool performs better given the type of data. So u should have some understanding of underlying methods. But straight away picking up the genes that are detected by all methods would not be a good idea. Still you could look at the common genes and also genes exclusively detected by a method, look at their expression profiles and get a sense of the data. But if you would like to select few candidate genes, then overlapping genes would be good choice.