I have a public scRNA dataset containing an expression matrix normalized by the deconvolution method in scran. The study sequenced two different types of cells; I am focusing on just one gene, and would like to see if it is differentially expressed between these two cell types. I am new to bioinformatics, and have only analyzed bulk RNA-seq data by DESeq2 before. I know there are similar programs for scRNA-seq, but they all seem to require raw count input, while I have a non-integer matrix. I have seen previous posts of people trying to input scran-normalized counts into MAST, and am wondering if this is recommended? Would it be bad practice to forgo the program and perform a Mann Whitney test on the normalized counts for my gene of interest instead?
Chapter 14 of the OSCA book covers DE in scRNA quite well. Chapter 13 may also be on interest. In short, if you had the original counts you could pseudo-bulk and perform a DE analysis via edgeR to see if the gene is differentially expressed. If you have an SCE or similar that contains those, this would be the route I would take.
Almost every DE method recommends raw counts and feeding any confounding factors to the GLM as part of the design formula. MAST does appear to work with normalized, log-scaled data, so you could give it a shot. If you post over on the Bioconductor support forum, one of the scran and/or edgeR authors will likely answer.
