DGE analysis in Seurat using paired samples per donor ?

Question

[Picasa]

Hi,

I have single-cell RNA-seq data from 5 donors, and for each donor, I have one Tumor and one Non-Tumor sample. I'm working with a Seurat object that contains all the cells, and I would like to perform a paired differential gene expression (DGE) analysis comparing Tumor vs Non-Tumor conditions while accounting for the paired design (i.e., donor effect).

Do you have an idea how can I perform this analysis using Seurat’s FindMarkers function?

Thanks in advance for your help!

Answer 1

[yura.grabovska]

If all you're interested in doing is doing tumour vs non-tumour from 5 cases then why not create a pseudobulk and just run normal DESeq2 on the resulting counts?

[Picasa]

I did already a pseudobulk approach. I would like now to compare with single cell specific methods (like MAST)

[yura.grabovska]

What do you think you'll gain by doing tumour vs normal at single-cell resolution? The whole point of single cell is to define sub-populations within the sample or group of samples. Because of the sparseness of single-cell data, you always lose power so it's important to understand why you're doing something. If you have a tumour and you're comparing it against a normal control without defining sub-populations then that's not really a single-cell experiment and you're just underpowering any results. Am I missing something?