Seurat VlnPlot vs PercentageFeatureSet expression level

Question

[Jeffrey3555054]

Hi I'm using the Seurat package for scRNA analysis and I'm trying to set a hard threshold to define cells as either high or low expression and was wondering is the expression level shown in VlnPlot() the same as the number obtained from PercentageFeatureSet() in the meta data? I also found out that the output for violin plot seems to change after I run PercentageFeatureSet(), this is the code I used:

VlnPlot(object, "MS4A1")

object[["MS4A1"]] = PercentageFeatureSet(object, "MS4A1")

VlnPlot(object, "MS4A1")

And the output for before is:

And after running PercentageFeatureSet() is:

Is there some sort of change in the data used to do the violin plot if I run certain functions to the seurat object? Thanks!

Answer 1

[rpolicastro]

In the first violin plot you're plotting the UMI count from one of the assay slots. In the second violin plot you created a meta-data column with the percentage of UMI counts MS4A1 represents compared to the total UMI counts for the cell, and are plotting that instead.

[Jeffrey3555054]

Does that mean the VlnPlot() function takes data from the meta data (if they exist) by default? And in this case, would using the percentage of UMI counts compared to total UMI counts be more reasonable than using the raw UMI count for comparing cells in different clusters?