Hi wondering why with identical settings other than EdgeR vs DeSeq2 would I get 5 orders of magnitude more diff peaks with EdgeR? Thanks.

There are a number of reasons why edgeR and DESeq2 may give different results. Five orders of magnitude is a lot however, how many consensus peaks do you have all together?

The most frequent cause of differences in my experience are in the normalization step. You can check how similar the normalizations are by generating MA plots for un-normalized, edgeR normalized (TMM), and DESeq2 normalized data. Here's a sample script that should work (and that you can adapt):

> data(tamoxifen_counts)
> tamoxifen <- dba.analyze(tamoxifen,method=DBA_ALL_METHODS)
> dba.show(tamoxifen,bContrasts=TRUE)
> par(mfrow=c(3,1))
> dba.plotMA(tamoxifen,th=0,bNormalized=FALSE,sub="Un-Normalized")
> dba.plotMA(tamoxifen, method=DBA_DESEQ2, sub="DESeq2")
> dba.plotMA(tamoxifen, method=DBA_EDGER, sub="edgeR")

You should be able to see that in this case, the DESeq2 analysis identifies twice as many DB sites, mostly with gains in the "Responsive" condition. The MA plots show that the TMM normalization in edgeR is adjusting more of the values "up" (so that most sites are centered around a fold change of zero), which reduces the detected differences. The more "conservative" normalization used for the DESeq2 analysis has a smaller upwards shift, preserving the concentrations of negative fold changes.

If this doesn't help, you can send me a link to your DBA object after the analysis (or after counting) and I can see if something else is going on. In this case please also email me the output of sessionInfo() so I can verify versions of other packages.

Cheers-

Rory

Thank you for the suggested script, using it currently to compare normalization. It is no longer 5 orders of magnitude, however it is two orders higher. I have about 94K consensus peaks. 

I guess Im just surprised to get 40-50 diff peaks from a consensus set of 94K. Using even higher FDR of .1

This may in fact relate to relatively stringent FDR set for the macs2 narrowPeak call.