I'm analyzing H3K27 peaks in two different cell populations and I want to get the differential peaks between the two using Diffbind. Peaks were called using MAC2 with the --broad flag. I run the analysis through DiffBind using the code:
K27<-dba(sampleSheet = "sampleSheet_K27.csv") K27<-dba.count(K27) K27<-dba.contrast(K27,categories = DBA_CONDITION, minMembers = 2) K27<-dba.analyze(K27, method = DBA_ALL_METHODS)
The analysis goes well but I noticed that there is a big difference between the differential regions identified by using EdgeR and DESeq2: none in the case of EdgeR and 1984 using DESeq2. I expect to see differences between the two methods principally because of their different normalization processes but this difference, in my humble opinion, seems to be too high to be caused only by normalization.
Do you have any idea where this difference may rely on? Is there something I do wrong? Or do I take the result from DESeq2?
Thanks in advance.
EDIT: below the MA plot as requested using the code:
> par(mfrow=c(3,1)) > dba.plotMA(K27,bNormalized=FALSE) > dba.plotMA(K27, method=DBA_DESEQ2) > dba.plotMA(K27, method=DBA_EDGER)