Estimating proportions of broad cell types

[rhiannonk]

Hello,

I am interested in using a finely annotated single-cell reference to estimate proportions of broader cell types.

For example, the Cell2Location tutorial uses a reference of 34 immune and non-immune subtypes.

If we wanted to estimate proportions for broader cell types (e.g., all B cells), would it be better to:

a) re-run the model with the broad cell type annotations, or
b) use the same model (with 34 subtypes) and sum the estimated proportions across relevant subtypes?

Curious to get any feedback on this issue!

-Rhiannon