Question about why non-overlapping peaks are being called reproducible

[l...@ucdavis.edu]

Hello,

I've been using IDR for TF ChIP-Seq analysis, and I've been pretty satisfied with the ability of IDR to identify reproducible and consistent peaks between biological replicates.

My brief understanding of IDR logic is that if a peak list is provided, IDR would identify all overlapping peaks in both bioreps and evaluate their reproducibility based on signal. Therefore, if a peak in the oracle peak file does not have a respective overlapping peak in both bioreps, it will be excluded from the final IDR list. 

However, there are some instances that IDR is considering peaks that are not present in one of the bioreps (see peak B in the example bellow). 

The code that I use for the analysis was the following:

idr --samples biorep1.peak.narrowPeak biorep2.peak.narrowPeak --peak-list oracle.peaks

Can someone please explain me why this is happening? Am I missing something with the logic?

Thank you very much for the help,

Best,

Leonardo