How to assign cell type labels to my 10X dataset using Allen Brain as a reference atlas?

I have 10X data from human prefrontal cortex, and I’d like to assign cell types. Using just a handful of marker genes, I can distinguish neurons from others in a UMAP plot, but I’d love a celltypist model or some other such tool. Does Allen Brain provide one?

I see there’s a well-defined nomenclature
and a paper describing it

but I can’t find a tutorial describing how to project my single cell data onto it.

Hi there! Thanks for your question. At this point, we don’t have any human prefrontal cortex data you can use to assign cell types. We are currently working on new reference datasets for PFC and MTG, along with a tool that will help you use them to annotate your own data, and will deploy these in the near future. We will follow up on this post when this is done.

In the meantime, we do have a 10X dataset from human primary motor cortex (M1) that might be a good starting point to help you annotate your data, described in this paper and highlighted in our Transcriptomics Explorer and Cell Type Knowledge Explorer. You can use this instance of the Azimuth tool to map the cell type calls from the M1 data onto your own data. Please note however that there are some discrepancies between the cell type labels in Azimuth and those represented in the links above, specifically that the Azimuth labels do not differentiate between Non-Neuronal and Non-Neural cell classes. We are working to get these labels updated in Azimuth and will follow up here when this is done.

As for the nomenclature standard you mentioned, if you are interested in applying this to your own data once you have annotated it, you can use the code provided at this GitHub repo.

Hope this helps!