differentially called peaks do not stay the same when inverting groups

[michellemeier27]

Hi there,

Thanks for developing such an awesome package and providing a detailed manual. I tried finding the answer to this question in the docs, but I think it might be very specific.

I've noticed that the results using the function getMarkerFeatures for pairwise comparisons is dependent on the order in which I feed them in. So, for instance cell type A vs cell type B gives me a different result than cell type B vs cell type A.
I'm a bit concerned because the differences are quite large and affect our interpretation of the data.
Just to give you some numbers, using cell type A vs cell type B, I get 20273 peaks more open in cell type A. Using cell type B vs cell type A, I get 21457 peaks more open in cell type A. Only 14996 of these peaks are shared. Would you have any explanations for how this happens?

Thanks for your help,
Michelle