Hello! I would like to start this by saying that I am new to Seurat and scRNA-seq. I would like to analyse the DE between patients. I have 6 patients (A1, A2, A3 and B1, B2, B3). I would like to analyse the DEG between A1 vs B1, A2 vs B2 and so on. I tried pseudobulking by following the Seurat vignette but ran into the following error:
> pseudo_markers <- FindMarkers(pseudo, ident.1 = "A1", ident.2 = "B1") %>%
arrange(desc(avg_log2FC))
Error in ValidateCellGroups(object = object, cells.1 = cells.1, cells.2 = cells.2, :
Cell group 1 has fewer than 3 cells
Does anyone have an idea how to fix this error and do pseudobulking?
•
written
by
bio_info •
Your setup isn't going to work as you'll only have one pseudobulk in each group you're trying to compare, which is insufficient to do differential expression. You could, however, do it group-wise with pseudobulks (e.g. A vs B).
That is likely what your error here is reflecting - lack of replicates. You should also be providing test.use = "DESeq2" to your FindMarkers call, as it's a method that works well on bulk samples.
This works perfectly! I played around with the metadata parameters and was able to get more replicates. Would using the Wilcoxon test be very different than DESeq2?
The Wilcoxon test needs more samples to be robust, so I'd trust DESeq2 on small numbers (<20) of pseudobulk populations more. As to how different they'd be, I'm not sure (though that'd be easy enough for you to test).
The Wilcoxon test is nice for when you're using the actual single cells as it's fast, simple, and the rank-based approach has been shown to have adequate power once you're comparing groups of sufficient size (e.g. dozens of cells in each) to each other. This old answer from the DESeq2 dev may be of some interest to you.
Log in to answer this question.
