How to assign cell types after integration in scRNA

Question

[Francesco]

Hi!

I need your help because I am a bit confuse about cell type assignment step after integrating different single cell RNA Seq datasets (2 conditions). Many integration tools in fact operate on a reduced dimensional representation and do not modify raw count or data layers, so when I use cell assignment tools like as azimuth, cellassign, celltypist (they operate on raw counts) I do not work on batch corrected data. What is, from your knowledge, the best approach to integrate different datasets and then assign cells?

Thank you very much

[ATpoint]

No difference here. After all you typically want a label per cluster/group, so either subset to the control condition to do the assignment, or use all cells. Subsetting probably makes sense if the conditions have strong perturbations that might not go well with markers based on homeostasis conditions. EIther do marker detection with control condition and compare to literature, or use automated classifiers given you have a reliable reference.

[Francesco]

Thanks for the fast reply.. however i had only partially understood the answer. Is correct saying that low-representation integration and batch effect correction is performed simply to align clusters between conditions? And after doing this, for degs calculation do i need to correct counts or not? Thank you very much

[ATpoint]

https://bioconductor.org/books/release/OSCA.multisample/using-corrected-values.html