How do I compare gene expression levels between clusters in single-cell RNA-seq dataset?

Question

[Sergei C]

Hi,

I am analyzing a single-cell RNA-seq dataset, in which expression values are log2(x+1) where x is gene counts divided by total counts within a cell.

In some downstream analyses, I found one cell-type cluster dominating a lot of signal.

Therefore, I wanted to characterize 1) whether this cell type had more expressed genes than other clusters within the same condition. 2) higher level of gene expression for a group of 800 genes than other clusters?

My data is just controls, so I am looking at cell type 1 vs cell type 2 .... vs cell type 17. I have 1500 cells and 17 cell types.

Thanks!

Answer 1

[ATpoint]

The term is 'differential expression', please see vignettes of OSCA (Bioconductor), Seurat or ScanPy. It's covered in great detail.

[Sergei C]

Thanks. The Seurat vignette linked this paper: https://www.nature.com/articles/nmeth.4612 . Looking at Figure 5, it appears that edgeR/QLF, MAST, and voom-limma performed better than other methods after filtering out lowly-expressed genes. I am assuming from differential expression testing that if one cell-type has many more upregulated genes compared to the other cell-types, it has a higher overall level of gene expression.

I was taking a look at Seurat's differential expression testing vignette. Would it be more logical to 1) find markers by inputting 17 cell_type identities FindMarkers(my_dataset, ident.1 = "cell_type1", ident.2 = "cell_type2", ...ident.17= "cell_type17", test.use = "MAST")

OR

2) Compare one cell-type to the rest of the dataset and repeat this 17 times for all cell types? FindMarkers(my_dataset, ident.1 = "cell_type1", test.use = "MAST")