Hello,

Looking for peace of mind really. We routinely perform bulk ATAC across multiple primary human cell types (fibros, immune cells). We recently did bulk ATAC-seq on human mammary epithelial cells and performed differential analysis with DESeq2. The design is 2 biological replicates of a 5 time-point time series. The counting step was done with summarizeOverlaps against an IDR-cleaned list of MACS3-detected peaks. We then contrasted each time point relative to the reference at Day 0. I have never seen this before, but DESeq2 only found mostly peaks with increased counts at one time point (Day 7), with close to 0 peaks with reduced counts. I have never seen such an unbalanced distribution (see MA plot and heatmap attached) . What is more interesting, is that when comparing against normalized signal tracks (by Deeptools), I can see several peaks with decreased counts relative to the Day 0 reference that DESeq2 did not catch (days going from top to bottom) . My padj cut-off is 0.05, which I think is still pretty relaxed. Has anyone seen something like this before? Not sure if I need to tweak something in the DESeq2 parameters or I should use something like featureCounts for the counting step.

Cheers for any insight,

# include your problematic code here with any corresponding output 
# please also include the results of running the following in an R session 

sessionInfo( )