Running celltypist with concatenated dataset

[malonzm1]

Hi,

I have datasets that I need to concatenate. Is it ok to run celltypist after concatenation or should I run it on the individual datasets before concatenation?

Thanks and good day.

Br,

[ChuanXu1]

@malonzm1, predicted_labels is not impacted, but majority_voting will be, due to the over-clustering step (different data structures will result in different clusterings).

[malonzm1]

@ChuanXu1 Thanks. Would you say that if I were interested in majority_voting it would be better to perform celltypist before concatenation?

[ChuanXu1]

@malonzm1, it depends. Usually, cells will be clustered in a better manifold when several datasets are combined and greater resolution is achieved, but sometimes clustering on a single dataset may reveal a better clustering shape (e.g., due to alleviated batch effect or optimal set of HVGs).