Issue with tl.annotate_cells_cluster function

Adding this issue on Petra's request.

My UMAP looks as expected (I had already run this analysis previously on besca24ngs)


The issue occurs when I label my celltypes and then re-plot the UMAP with the labels. I get the following gibberish. I know that the labelings are good.


The problem was sorted by using a different labelling function as described by Petra.

# Dictionary for renaming values
correspondence = {
'0': 'Tumor_Necro',
'1': 'Tumor',
'2': 'Tumor_Necro',
'3': 'Tumor_Fibro',
...
}

adata.obs['annov2'] = adata.obs['leiden'].copy()
adata.obs['annov2'] = adata.obs['annov2'].replace(correspondence)

So seems like an issue with the tl.annotate_cells_cluster function

Start by reproducing the reported behavior around tl.annotate_cells_cluster, using the described UMAP labeling workflow and the adata.obs values shown in the issue. Compare its output with the replace-based workaround. Done means the annotated labels render correctly when the UMAP is re-plotted, with a regression check for the reported labeling case.

Written by the indexing model from the issue text.