Cell2location workflow, normalized for abundance?

[Dillon214]

Hello ncem devs,

I have a question about how to use cell2location celltype-specific counts (generated by the compute_expected_per_cell_type method) for NCEM. I know there is a workflow available, but I wanted to ask a question regarding it...

Prior to ncem, are celltype-specific counts generated by cell2location normalized for abundance? I ask this because cell2location devs have suggested both pruning and normalizing cells based on abundance prior to feeding the resulting matrix into ncem. See link and attached image: BayraktarLab/cell2location#211.

Do you reccomend this procedure as well, and if so, is it included in the available workflow. If it is not included, what kind of normalization procedure would you recommend?

-Dillon