scRNAseq for mutant and wt does not reveal differential expression of knocked-down gene

Question

[samuel.storey]

I have four scRNAseq datasets we prepared consisting of whole-retina samples for each. Half of these datasets are wildtype and half are a transgenic knock-down for a specific gene.

My seurat differential expression analysis does not reveal our mutant gene as a differentially expressed gene between the WT and the mutant. I am able to find mRNA expression in both WT and mutant, however a comparative FeaturePlot analysis would suggest that there is far less mRNA expression of our gene in the mutant. We've attributed this to nonsense mediated decay, and is well supported by our qPCR runs.

I've integrated all four datasets, and I was initially relieved to see that the gene in question was in my RNA assay and not the integrated assay which I had stupidly defaulted to for the initial differential expression analysis. However, even fixing this error, the gene in question is not observed in differential expression.

I am following this tutorial for the differential expression but I am wondering if perhaps anyone has any ideas for why this might be? Is there some consideration one must take when integrating and then subsequently performing the differential expression? For certain cell types the FindMarkers output between conditions is very few (eg. four differentially expressed genes out of thousands).

Any advice would be appreciated,

S.

[ATpoint]

Please show any data.