Merge or Integrate Non-Overlapping Datasets - Different Experiments and Cell Types

Question

[kebl4660]

I want to run a joint-analysis on publicly available scRNA-seq of different cell types say Monocytes, T Cells and Neutrophils. If I have 3 different datasets, one for each population, is it better to MERGE them or INTEGRATE them? On one hand, I would like to account for batch effects and technical variations. On the other, Seurat states the integration method requires at least some overlap between cell types which is not my case. Can I still run the integration workflow? Will it give me meaningful results? Or should I stick to merging the datasets?

Thank you!

[rpolicastro]

If each dataset has non-overlapping cell types it wouldn't make much sense to integrate them. Integration is designed to "smooth out" the differences in samples effected by batch effects so that cells that are likely of the same cell type cluster together. Depending on what your ultimate goal of this analysis is I may recommend avoiding this joint analysis, but without any more information I can't give any specific recommendation.