SCT-residual vs corrected counts in metacell and coexpression analysis

[xyang2uchicago]

Hi, SCT normalization group,

I have been confused about which is the proper input for my metacell analysis and gene coexpression estimation.
After searching the Q&A here, I see the recommendation to use corrected counts for DE analysis but the residuals for PCA, reduce dimension, and neighboring analysis. While I did not find discussion on the proper input for new metacell (such as SEAcell https://www.nature.com/articles/s41587-023-01716-9) analysis. Neither for coexpression analysis.

Can anyone put theoretical insights or share practical experience here?

Thanks
Holly