When I run diffbind on just 2 groups in my experiment, I get 31,738 consensus peaks and the DESEQ2 analysis looks beautiful in IGV. But then when I run diffbind on all of the groups in my experiment together (10 groups total), I lose some of those peaks in the new consensus peak set. I have tried the following things, none of which have retained all of the peaks when I run just the 2 groups on their own:
- Using the minOverlap parameter and turning off summits:
dbObj <- dba.count(dbObj, bUseSummarizeOverlaps=TRUE, summits=FALSE, minOverlap=0)
- Using the consensus-of-consensus approach:
dbObj.cons<-dba.peakset(dbObj,consensus = DBA_FACTOR, minOverlap = 0) dbObj.cons<-dba(dbObj.cons,mask=dbObj.cons$masks$Consensus) consensus.peaks<-dba.peakset(dbObj.cons,bRetrieve=TRUE) dbObj$config$singleEnd <- FALSE dbObj <- dba.count(dbObj,bUseSummarizeOverlaps=TRUE,summits=FALSE, minOverlap=0, peaks = consensus.peaks)
Any idea what could be going on?