Hi everyone, I am analyzing super-enhancers (SEs) from H3K27ac CUT&Tag data and have observed some confusing results. I would greatly appreciate any advice on my workflow and result interpretation. I have three biological replicates per condition. My original pipeline was: call peaks individually for each replicate with MACS2, merge peak files using bedtools merge, and pool all replicate BAM files for downstream ROSE SE calling.

I noticed several unusual features in my SE results, and I'm unsure whether these are normal or technical artifacts:

1. SE regions appear relatively broad: average ~30 kb and a maximum of ~180 kb.

2. Over 40% of SEs annotate to promoter regions (<1 kb from TSS).

I suspect that these unexpected patterns may result from my simple replicate merging strategy, which might retain noisy, non-reproducible peaks and artificially expand SE intervals. I have two key questions:

1. Are these metrics (broad SE regions, high promoter occupancy) acceptable for standard SE profiles, or are they likely preprocessing artifacts?
2. What is the proper, robust strategy to integrate three replicates for ROSE? Should I use a consensus peak set (2/3 replicates overlap) instead of simple merging to filter noise and avoid over-broad SE calls? Thanks in advance for any suggestions or shared experience!