Hi all,
I am looking at the expression of a bunch of genes of interest in scRNA-seq. The dataset has two conditions to compare: pallial and subpallial. When using the DotPlot function of Seurat (cf dotplot screenshot), it seems like gene "Smo" is more expressed in pallial than in subpallial.
DotPlot(seurat_object, features=genes, group.by="region", scale = TRUE)
When plotting with VlnPlot of Seurat (cf violin plot screenshot), it looks like the difference is not that big.
VlnPlot(seurat_object,features="Smo", split.by="region", layer="data")
Also, I use FindMarkers as follows to see if the gene expression difference is significant:
FindMarkers(seurat_object, ident.1="pallial", ident.2="subpallial", features="Smo")
-> it does not pass the logfc_threshold of 0.1 : does it mean the difference in expression is super small between the two conditions?
And I try a wilcoxon statistical test as follows (where region encodes for "pallial" and "subpallial" information):
df <- FetchData(qNSC_object, vars=c("Smo", "region"))
wilcox.test(Smo ~ region, data = df)
-> it is not significant (p-value=0.7917) (while it's at the cell level, it's not even a pseudobulk DE taking into consideration batches, so we would rather expect that it inflates significance - in other words it must really not be a significant difference)
I therefore suspect the DotPlot to exaggerate the difference in Smo expression between pallial and subpallial. What do you think? I also find it strange to have positive expression in one condition and mirrored negative expression in the other condition. Is it some kind of z-score centered around zero? I also want to make sure that it uses the logged-normalized value of the "data" layer of my Seurat object, isn't it?
If anyone would know how to properly compare those two conditions in terms of "Smo" gene expression it would be awesome!!
Thank you so so much :)
