Fixed peak width in ENCODE TFs ChIP-seq

Question

[DavideBrex]

Dear all,

I have a question about the TFs ChIP-seq peak width of the ENCODE project.

It is probably something I am missing. If I understood correctly, ENCODE TFs ChIP-seq pipeline (see here) uses SSP for peak calling since it was better for IDR reproducibility (check GitHub issue here).

If I look at the distribution of peak length for a random TF on encode (bigBed narrowPeak - optimal IDR thresholded peaks), I see this distribution:

I checked the SPP paper, but I couldn't figure out why we see an enrichment of peaks, all with the same length and some with other lengths.

This does not occur with MACS2.

What am I missing? Thank you!