Recommendations for filtering scATAC-seq pseudobulk data

[blake-bowen]

Thanks for the great tool and guide for best practices in scATAC-seq differential accessibility analysis.

I was wondering, do you have any recommendations for filtering pseudobulk profiles prior to identifying differentially accessible regions with edgeR or DESeq2.

Specifically:

1. Filtering out pseudo-bulked samples that do not have very many cells or low library size
2. Filtering out regions that are not accessible in the majority of samples

When performing DA analysis libra do you typically do either of the above and if so, what minimum thresholds / rules of thumb would you recommend?

Thank you!