It would be really great if you could somehow make the function sca.models.SCANVI.prepare_query_anndata able to take into consideration both gene Ensembl IDs and gene symbols (i.e. actual gene names), in particular for the subsetting and padding that is done there. Do you think that would be doable?
For example, the Human Lung Cell Atlas model was trained on gene IDs, but some datasets might only have gene names available. As I have the mapping of the gene IDs to gene names available for the HLCA model, it should be relatively easy to accommodate for those datasets.
Hi!
It would be really great if you could somehow make the function
sca.models.SCANVI.prepare_query_anndata
able to take into consideration both gene Ensembl IDs and gene symbols (i.e. actual gene names), in particular for the subsetting and padding that is done there. Do you think that would be doable? For example, the Human Lung Cell Atlas model was trained on gene IDs, but some datasets might only have gene names available. As I have the mapping of the gene IDs to gene names available for the HLCA model, it should be relatively easy to accommodate for those datasets.