Filter ATAC before running scDblFinder

[ccruizm]

Good day,

I wanna use scDblFinder on my scATAC data. For RNA, you warn users to perform initial QC so it does not influence the modeling of doublets for more precise doublet calling.

I have checked the ATAC vignette but could not find information about this particular point. What would you recommend?

I appreciate any suggestions you can give me.

[plger]

Hi,

Actually it turns out that scDblFinder and its thresholding procedure are pretty robust to prior filtering or not, so I don't think the results will change much (unless you're filtering out a majority of cells).
The same therefore applies for ATAC-seq when using the aggregation-based method.

Regarding the Amulet method, it's anyway is largely independent across cells and hence of prior filtering (the only impact across cells is the filtering out of possible repeat regions).

So in conclusion: it shouldn't make much of a difference.

[ccruizm]

Thanks for the speedy reply and for clarifying this! It is good to know that it won't impact the performance of the doublet calling.