Hi community,

I was trying to use DESeq2 to analyse pseudobulk single-cell RNAseq data. Someone else already made the cell type identification for me, so I have combined counts of each treatment from each cell type.

As opposed to other "treatments", there's no "mock" cell type in the data, and I'm interested in how cell types affect the response to my treatments. If I just have the standard ~ cell_type*treatment model, one cell type will be set as reference and all the cell_type:treatment interaction terms will be showing the additional effect compared to that reference cell type. I'd rather like to compare the cell type-specific response to the average response. Is it correct to make an additional "mock" cell type by adding up the counts of all cell types from the mock treatment condition? I suppose DESeq2 will normalise for the library size right?

Thanks in advance for any input!

The coding of the interaction terms doesn't matter for a LRT (the fact there is an arbitrary reference cell type), you would just compare the two models.