Tuning seurat integration strength to prevent overcorrection

Question

[nhaus]

Hello,

I am currently working with multiple replicates for 2 different conditions (stimulated and unstimulated cells) that I want to integrate to adjust to correct for batch effects.

I was wondering, if there is a way (or parameter), which can be adjusted to control the integration strength or stringency. I am afraid that cell types that are unique to the stimulated condition, are projected onto the "wrong" cells, because the integration is to stringent, so I want to try out different settings.

Any help is much appreciated.

Cheers!

(I also asked this question on the github issue page of seurat)

Answer 1

[fracarb8]

You need to clarify which type of integration you are performing (rPCA or CCA?). If it is rPCA, you can tweak the k.anchor argument. The seurat documentation has a section dedicated exactly to that: Modifying the strength of integration.

[nhaus]

Unfortunately, I am using the CCA based approach.

[fracarb8]

It should work for CCA as the functions (FindIntegrationAnchors) are the same.

[nhaus]

Good point! Thank you very much. I will try it out and report back!