Differential Accessibility

[markphillippebworth]

We need to add a function to calculate differentially accessible regions.

Ideally, it would use sample-specific information to find differential peaks, and
then iteratively downsample those peaks to determine if they consistently show up as differential or if the difference arises from drop-out effects at lower cell numbers.

[markphillippebworth]

Addendum: I'm think we should actually normalize by reads in some way. Normalizing by cell number will work as long as the sequencing depth between the two samples. If the sequencing depth changes, then so will the odds of having a drop-out event. In that situation, both cell number and sequencing depth together will influence the drop-out rate and so both need to be controlled for.