I have an old post describing some noisy ATAC-seq data. I went back and did some more diagnoses. However, I have no idea why the results are so different from my expectation. Therefore, would like to seek advice from the community on:

1. What is the issue of the ATAC-seq library? Is it library preparation error, sequencing error, or wrong analysis pipeline used?
2. Is it worth conducting downstream analysis on this library preparation? (I did a crude analysis on DiffBind using all default setting, and found no differences by DESeq2 while 8 significant results from EdgeR)

Here are some diagnostic plots you may found it useful:

I got around 200000 peaks for each sample. And here is the some of the typical TSS enrichment plot:

I found that there is an enrichment of fragment at around 150-200bp:

However, I found that this might be actually normal for frozen mouse liver tissue (the tissue I used) from a scientific data publication:

The same "contamination" can also be visualized in M-plot:

I checked the peaks around Gapdh locus and found that seems the data is contaminated with noise.

Thanks in advance!