integrated or merged data?

[algarji]

I was wondering if this CrossTalkeR pipeline could be applied in merged data from two or more 10x runs.

I think that this makes more sense in integrated data, where cells with shared similar global identity are grouped together despite their differences, and so cellphonedb or CrossTalkeR can further explore LR interactions per condition.

However, in my case, integration is not an option as it mixes populations of cells that are important to be separated in different clusters, so I directly merged data. With merged data, clusters of cells are going to be very similar regardless condition (and differences in cell count per cluster can be also important between conditions). Despite this, when I use CrossTalker I can see enrichment in some LR interactions.

Do you think that CrossTalkeR can be performed with merged data? If not, maybe a solution could be to integrate data just to follow this pipeline, or maybe to take very general clusters (e.g. NK + cd8_eff + gd --> cytotoxic), and so the effect of LR interactions between conditions will be easier to assess.