Highest variable features in single cell data

Question

[carolofharvest]

I used the 'FindVariableFeatures' function from the Seurat package to identify variable features, but some of the genes appearing in the results are only expressed in 10-15 cells, and these cells are not even in a single cluster. What should I do in this situation?

[ATpoint]

and these cells are not even in a single cluster.

What does this mean? If you run clustering on all cells then every cell is assigned to one cluster. Check if strange cells are outliers in a QC metric. Or it's just a celltype that is not abundant or poorly captured by the single-cell tech.

[carolofharvest]

I apologize for the confusion. What I meant to say is that, for example, the Trbv17 gene appears among the variable genes. However, when I plot the feature plot, this gene is expressed in only a very small number of cells, and these cells are scattered across random clusters on the UMAP plot.

Compared to the total number of cells, the number of cells expressing Trbv17 is very very small. I don't understand how this gene can be detected as a highly variable gene in this case since majority of the cells dont express this gene.

[bk11]

You need to be clear about how Seurat defines highly variable genes here. Highly variable genes are the genes that have very high expression in some cells and low or no-expression in other cells. Thus in your case, Trbv17 gene is rightly picked as a variable gene as you are seeing in your featureplot its expression in a very few cells, which is totally expected. There is nothing wrong with it.