RunNMF with SCTransform

[chrarnold]

I have a question for RunNMF for Seurat objects from Visium data which have been normalized using SCTransform. The default slot name for the function is scaled.data. SCTransform stores the normalized data in the counts slot, however, and the Pearson coefficients in the scaled.data slot. I feel using the default scaled.data slot is incorrect in the case, could someone comment on it? Thank you.

[chrarnold]

This is described here, for example: https://satijalab.org/seurat/articles/sctransform_vignette.html. I am just not sure whether running RunNMF with the residuals rather than the corrected counts is the right thing to do.