"ZINBA was run using a window size of 300 bp and an offset of 75 bp. Alignability was used to model the zero-inflated component, and the ATAC- seq read count for the background and enriched components. Enriched regions were identified as those with a posterior probability of >0.8" \cite{Buenrostro2013}
Fragments up to 2 kb allowed to align (-X2000) and only unique aligning reads were collected (-m1)
Reads aligning to the + strand offset by +4 bp, and reads aligning to the – strand offset −5 bp.
Call peaks with Zinba: