Best way to do L-R analysis across separate tissues

[pbhatia91]

Hi! Thanks for developing this awesome tool!

I want to L-R analysis across two different tissue types, i.e., they don't have the same cell type clusters and and I want to see the L-R interactions between the cell types of one tissue across the other tissue.

The initial way I thought of doing it is just merging the Seurat objects but I am not sure whether I should use re-normalized data across both tissues or keep the original normalization from their respective objects and then run it with LIANA. Do you think the latter is the right way to do it? Your advice on this would be appreciated! Thanks!

[dbdimitrov]

Hi,

Thanks for the kind issue.

I'm not familiar with your data and experimental design, but I potentially have some concerns associated with the potential batch when inferring communication events from different experiments? and tissues. I would be careful with those.

To answer your initial question, if we assume that this is not an issue and interactions between the tissues are plausible, then you should have comparable cell libraries, i.e. you should indeed re-normalize the data with all atlases together.

Hope this helps.