I have a DE matrix of Single cell RNAseq data for two conditions. there are 500 genes that has FDRp<0.01, FC>2. I want to reduce the list to say around 100-200 genes to make it more presentable. Is there any stringency criteria that I can use for SCRNAseq data, like sequencing depth, read counts? what is commonly used for such purposes in ScRNAseq data.
I did not work a lot in scRNA but general thoughts.
It largely depends on which method you used for differential analysis. It's common to rank differential genes by Adj.p-value or by fold-change.
For scRNA, as you would be having many data points, ranking by differences in the mean expression between two conditions or cell-types would also pick-up highly differential genes. As scRNA is sparse, I would also check proportion of cells expressing those genes.
You could also do clustering on DE matrix and check if you get clusters (genes) that show very distinct patterns in two conditions/cell-types.
There are also logistic regression approaches to pick-up top 'n' genes that could accurately classify cell-types or two conditions.