FindAllMarkers()

[rythi4]

I'm running FindAllMarkers() on a published dataset, where I downloaded the metadata and a normalized count matrix and made a seuratobject.
I'm trying to use FindAllMarkers() to find positive markers, but I don't understand the results I get:

I don't understand how pct.1 can be lower than pct.2 and still be a positive marker, with a positive log2fold change.

I ran:
Markers <- FindAllMarkers(seurat_object, min.pct = 0.25, only.pos = TRUE)

[samuel-marsh]

Hi,

I believe only.pos refers to the log fold change values and not the percent expressed values. So it appears to be functioning as expected. You can always filter the returned dataframe after running FindAllMarkers for anything you want, including whether to only include samples that have higher pct.1.

Best,
Sam

[rythi4]

As I understand it, avg_log2FC is a comparison between the expression in the two groups. And therefore positive values indicate that there is more expression in group 1, and negative values indicate that there is more expression in group 2. Therefore, my initial thought was that when avg_log2FC is positive, then pct.1 > pct.2.
But could the avg_log2FC be positive when pct.1 < pct.2, because the few cells in group 1 have very high expression of the gene in question?

[crazyxiaoj]

I also encountered a similar problem. I looked at the violin plot and was surprised to find that pct.2 has a larger "belly" compared to pct.1, and it seems that pct.2 has higher expression than pct.1. But then I looked at the dot plot and found that the dot plot showed a trend of higher expression in pct.2, while the trend in the violin plot was wrong. So you can try using a dot plot for confirmation?