How to change FeaturePlot expression scale to z-score transformed scale

[ksaunders73]

Hello!

This may be because I do not fully understand the source code for FeaturePlot() (https://github.com/satijalab/seurat/blob/master/man/FeaturePlot.Rd), but by what I think I understand, the scale on the righthand side is in terms of relative expression of a feature to the maximum expression of that feature (to the whole dataset or to the clusters or to specified conditions if they are specified?).

Is there a way to transform this relative expression into a Z score (or some other statistic) and make the scale show these Z scores?

I am also running some code from #3521 which allows me to plot a list of genes onto a single Feature Plot:

nkfocuslist <- list(c("TFF1", "MB", "ANKRD30B",
             "LINC00173", "DSCAM-AS1", "IGHG1", "SERPINA5"))
sobj <- AddModuleScore(object = sobj, features = nkfocuslist, name = "NK_Focus_List")
FeaturePlot(object = sobj, features = "NK_Focus_List1")

Is there a way to make this scale a Z-score scale as well?

Hopefully this makes sense, and thank you for reading!

[yuhanH]

Hi,
This ModuleScore looks a little wired in your plot. Have you run NormalizeData before AddModuleScore?
Since module score stores in the meta.data, you can convert it to the Z-score directly.

sobj$NK_Focus_List1 <- (sobj$NK_Focus_List1  - mean(sobj$NK_Focus_List1 )) / sd(sobj$NK_Focus_List1 )

[ksaunders73]

Hello @yuhanH ! This worked! Thanks so much for your help!! As for my data, it is a TPM matrix so I skipped the normalization step as I've seen recommended on this forum (#668) and the bioinformatics stack exchange (https://bioinformatics.stackexchange.com/questions/5115/seurat-with-normalized-count-matrix). I only ran FindVariableFeatures() (ExpMean), ScaleData() (default linear regression), and then RunPCA(), FindNeighbors(), FindClusters(), and then the RunUMAP().