Unexpected number of genes identified with FindVariableFeatures()

[EliseCoopman]

Hi,

I am using the FindVariableFeatures() function from Seurat to identify genes with the most variability in our dataset. I'm using the default parameters (e.g., nfeatures=2000), but I end up identifying only 656 genes, which is much fewer than expected.

Is it correct that FindVariableFeatures() calculates variable genes on an individual basis (for each counts layer) and then aggregates them into an overall list? How does Seurat decide to end up with just 656 genes? Is this behavior expected, or might there be an issue with how variability is being calculated or selected? Here the code I used:

library(BPCells)
library(Seurat)

seuratobj_splitted <- SplitObject(seuratobj, split.by = "sample_id")

# Remove lowly expressed genes
seuratobj_splitted <- lapply(seuratobj_splitted, function(x) {
    expressed_genes <- rowSums(x[["RNA"]]$counts > 0) >= 10 
    expressed_genes <- names(expressed_genes)[expressed_genes] 
    subset(x, features = expressed_genes)
})

for (lib in names(seuratobj_splitted)) {
    seuratobj_splitted[[lib]]@project.name <- lib
}

# Merge into a single object
seuratobj <- merge(x = seuratobj_splitted[[1]], y = seuratobj_splitted[-1])

seuratobj <- NormalizeData(seuratobj)
seuratobj <- FindVariableFeatures(seuratobj, selection.method = "vst", nfeatures = 2000)
#Finding variable features for layer counts.individualA
#Finding variable features for layer counts.individualB
#Finding variable features for layer counts.individualC
#Finding variable features for layer counts.individualD
#...

top2000 <- head(VariableFeatures(seuratobj), 2000)
length(top2000)
[1] 656

Thanks!

Elise