Detection of unknown sub cells in a population of cells using DEGs from scRNAseq data

Question

[ashkan]

in a scRNAseq data set, I have a list of DE genes and want to use them to detect unknown sub cells in the population of cells in the same dataset. how can I do it? I am using Seurat package.

Answer 1

[ATpoint]

Given that these DEGs form a geneset that is enriched or specific in your "sub cells" you simply run clustering (with parameters set go get a lot of small clusters) and then simply color clusters by expression of these genes. Either the geneset as a score, or individual genes, and see whether one or some clusters are indeed convincingly expressing these genes highly. No magic here I would say.

[ashkan]

ATpoint what if we have few genes and want to isolate the cells that express all those genes. basically we know there should be a small population of cells in our dataset which can be isolated in FACS. I want to take those cells. in my scRNAseq data when I do DEG analysis and make the dot plot, I will ge few DEGs but based on the dotplot, they are not expressed in the proportion of cells. I want to isolate the cells that have those genes in comon.

[ATpoint]

I stand by what I said, calculate a signature score per cell (UCell package for example) and see whether you find cells that qualify to represent this population.