Should we calculate outliers before or after removing doublets?

[derrik-gratz]

Our current workflow looks roughly like

Doublet detection Outlier detection / cell filtering Clustering Remove doublets

It might be better to do

Doublet detection Clustering Remove doublets Outlier detection/cell filtering Clustering (agian)

That way were doing outlier detection on singlets, or drops we decided to move forward with.

Note that scDblFinder does not want pre-processed data as an input, so it should be prior to any other filtering besides empty drops (source)

[micahpf]

Ah that's a good point. I think this workflow is leftover from when we were using doubletfinder, which wanted a qc filtered object.

[derrik-gratz]

Thanks for checking, I'll work on modifying the workflow