Is there a way to find out which reference genes drive the cell type prediction score?

[dgoekbuget]

In an attempt to make the cell type prediction score more biologically interpretable, I am wondering if you could suggest a way to say identify the top 10 genes driving the score for a particular cell type? Those could then be (feature)plotted and shown to be specific in the query population of interest. Thanks for a great tool.

[erzakiev]

That is a good request, I double that