What is the SnapATAC2 definition of a "reproducible" peak?

[adamklie]

Hey Kai,

Hoping to get some details on how exactly reproducible peaks are determined when using the replicate argument of snapatac2.tl.macs3. I'm familiar with the idea of calling peaks in each replicate and in fragments pooled across replicates, and then only keeping peaks from the pooled set that are reproducible in the replicates. But what exactly is the definition of "reproducible" here. Does the peak need to overlap in a single replicate? All the replicates? Is overlap considered to be 1bp, 50% of the peak etc.?

Thanks again for all the help!

[kaizhang]

Current definition is that the reproducible peak need to overlap with all replicates with at least 1bp. But I can expose these parameters to the user end.

[adamklie]

Might be useful to be able to fiddle with those parameters. Peak calling feels like more of an art than a science sometimes.