Wilcox test using data slot (log normalized count values) of scRNA-seq between two groups

Question

[mropri]

Hi,

I was wondering would one be able to do a wilcox test for a specific gene in the same cell type in two different conditions using the log normalized values in seurat (the data slot of the RNA assay)? For example, if I have the gene EGFR, and I want to see if there is a difference in expression of EGFR in epithelial cell of a control sample/samples vs the epithelial cells of a treatment sample/samples. I know I can run edgeR or DESeq2 for differential expression for full genes but would a wilcox test also work, especially if I want to test the difference for a single gene? Appreciate any insight.

Answer 1

[yxwucq]

The best way is to generate a pseudo-bulk aggregation then apply edgeR or DESeq2. You can also directly use wilcoxon on sc-level, since many people are doing this, but be aware of possible false-positive results.

You can find details in this post: Difference between GLM and Wilcoxon rank sum test

[mropri]

Thank you, will do Wilcox for now as I only have one control and one treatment, as I get more control and treatment samples, will stick with pseudobulk and DESeq2.