We have a 2-factor experiment. Factor-1 is pre vs post treatment. Factor-2 is cured vs not-cured. When performing edgeR I input factors as pre, post, pre, post, etc. And lib-types were cured, cured, not-cured, not-cured, etc. I imputed "genes" into my counts file as follows: marker pre/post up (all pre samples had read counts of 200, all post samples had read counts of 2,000); marker pre/post down (reverse of the above); marker cure up (all pre samples in cured animals had read counts of 200 and post samples in cured animals had read counts of 2,000 - for all not-cured animals both pre and post had counts of 200); marker cured down (all pre samples in cured animals had read counts of 2,000 and all post samples in cured animals had read counts 200 - all samples pre and post in not-cured animals had read counts 200). EdgeR correctly identified my markers for cure/not-cure as significantly DE and the markers for pre/post were not significant. BUT, the log FC for the cure/not-cure markers (marker up and marker down) were BOTH NEGATIVE. So what am I doing wrong? EdgeR is not finding the pattern we are looking for. Namely, significant change between pre and post in cured but no significant change in pre vs post in not-cured.
Am I doing the factors and lib-types incorrectly?
Put some code in describing what you did, a big block of text isn't very easy to read or understand.
Thanks for the response. Unfortunately I'm using RNASeqGUI. Basically a GUI for RNASeq data analysis. So I have no code to give you. My only imputs are the fields, factors and lib-types. So factors are pre,post,pre,post,etc. and lib types are cured,cured,not-cured,not-cured,etc For 34 animals. And of course I input the counts file. Should I put something else for lib types or factors?
RNASeqGUI isn't a Bioconductor package, so you'll have to ask the authors about what's happening inside their software. We can't make any statement based on what they may or may not be doing.