DE section issues

[le-ander]

I'm just looking at the DE section of the best practice book and noticed two potential issues with the aggregate_and_filter() function:

The text says "For each patient we create 1 pseudobulk sample per cell type by aggregating the cell from each subpopulation and taking the mean gene expression within that subpopulation." but in the function, counts are summed rather than mean aggregated. Additionally, it might be helpful to add any required columns (eg. "donor_key", "condition_key" etc) to the 'obs_to_keep' list to prevent an error where the donor_key is not found in the donor_df

[alitinet]

Hi, should be fixed after we move to decoupler (#141).