Why are some cells of Cell type A located in Cell type B in scRNA-seq clustering?

Question

[joonhong kwon]

Hi all,

I have a question about the scRNA-seq clustering. In my data or some paper, although Cell type A and B are different cell types and are located far away in UMAP, some cells of Cell type A located in Cell type B. Take my UMAP as an example, Astro4 (green) clusterd well with other Astrocyte cluster, but some cells of Astro4 located in Microglia clusters (topright).

I think this is caused by a doublet issue or low capture efficiency or something...but I'm not sure...

How best to explain this phenomenon?

Thank you,

Joonhong

[ATpoint]

Clustering is commonly done on the principal components while UMAP is a different algorithm, so some discrepancy is not unexpected. It would also depend on what the colors are based on, is this clusters as in the graph-based clustering, or something else? HArd to comment without details.

[joonhong kwon]

I performed the analysis by following the seurat integration tutorial. This method is graph-based clustering.

That figure show only astrocyte clusters among all brain cells using subset function.

https://satijalab.org/seurat/articles/integration_introduction.html

Thank you.