RuntimeError when creating custom reference

Hello @danielsf,

Thank you for putting together this wonderful tool. For our analysis workflow, we would like to be able to create region specific reference based on the ABC WMB taxonomy. I followed the instructions to create the precomputed_stats.h5 file from a ABC release as documented here.

The inputs for the function include all of the original taxonomy CSV files, the region specific cell metadata file and region specific list of h5ad files. However, I am running into a RuntimeError raised by the validate_taxonomy_tree function.

RuntimeError: node CS20230722_CLUS_0001 (child of CCN20230722_SUPT:CS20230722_SUPT_0001 -- type <class 'str'>) is not present in the keys at level CCN20230722_CLUS e.g. of keys ['CS20230722_CLUS_0326', 'CS20230722_CLUS_0328', 'CS20230722_CLUS_0033', 'CS20230722_CLUS_0038', 'CS20230722_CLUS_0044']

I separately confirmed that the CSV file includes all the parent-child relationships and that level CCN20230722_CLUS does include the node CS20230722_CLUS_0001. I am a bit confused about the setup since the above function is getting the taxonomy tree input from the h5ad file which for the ABC releases are devoid of taxonomy information. Is there a work around for this error ? Thank you for your help!

Start with docs/input_data_files/precomputed_stats_file.md and inspect validate_taxonomy_tree in src/cell_type_mapper/taxonomy/utils.py. Reproduce the RuntimeError using the reported ABC release inputs, then trace how taxonomy data is obtained from the h5ad files and CSV files. Done means identifying and documenting or fixing the mismatch so region-specific reference creation validates successfully.