showing gene expression on umap not working

When I tried to plot the expression of a particular gene on umap map by the tutorial, it always showed the following error:

>>> sc.pl.umap(post_adata, color=['XKR4'])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:\ProgramData\Miniconda3\lib\site-packages\scanpy\plotting\_tools\scatterplots.py", line 542, in umap
    return embedding(adata, 'umap', **kwargs)
  File "C:\ProgramData\Miniconda3\lib\site-packages\scanpy\plotting\_tools\scatterplots.py", line 207, in embedding
    use_raw=use_raw, gene_symbols=gene_symbols,
  File "C:\ProgramData\Miniconda3\lib\site-packages\scanpy\plotting\_tools\scatterplots.py", line 865, in _get_color_values
    values = adata.raw.obs_vector(value_to_plot)
  File "C:\ProgramData\Miniconda3\lib\site-packages\anndata\core\anndata.py", line 413, in obs_vector
    idx = self._normalize_indices((slice(None), k))
  File "C:\ProgramData\Miniconda3\lib\site-packages\anndata\core\anndata.py", line 364, in _normalize_indices
    var = _normalize_index(var, self.var_names)
  File "C:\ProgramData\Miniconda3\lib\site-packages\anndata\core\anndata.py", line 155, in _normalize_index
    return name_idx(index)
  File "C:\ProgramData\Miniconda3\lib\site-packages\anndata\core\anndata.py", line 142, in name_idx
    .format(i))
IndexError: Key "XKR4" is not valid observation/variable name/index.

However, the gene XKR4 did exist in the var_names:

>>> post_adata.var_names
Index(['XKR4', 'RP1', 'SOX17', 'MRPL15', 'LYPLA1', 'TCEA1', 'RGS20', 'ATP6V1H',
       'OPRK1', 'NPBWR1',
       ...
       '2700089I24RIK', 'RAB11FIP2', 'E330013P04RIK', 'NANOS1', 'EIF3A',
       'FAM45A', 'SFXN4', 'PRDX3', 'GRK5', 'CSF2RA'],
      dtype='object', length=16249)

The anndata object looked as below and it was fine when I tried to show the louvain clusters:

>>> post_adata
AnnData object with n_obs × n_vars = 88291 × 16249
    obs: 'CellID', 'batch_indices', 'labels', 'local_means', 'local_vars', 'louvain', 'clusters'
    var: 'gene_id'
    uns: 'neighbors', 'louvain', 'louvain_colors'
    obsm: 'X_scVI', 'X_umap'

>>> sc.pl.umap(post_adata, color=['louvain'])

...

Versions:

scanpy==1.4.5.post3 anndata==0.6.22.post1 umap==0.3.10 numpy==1.18.1 scipy==1.3.2 pandas==0.25.3 scikit-learn==0.22.1 statsmodels==0.11.0 python-igraph==0.7.1+5.3b99dbf6 louvain==0.6.1

[aditisk]

@LuckyMD thanks a lot for the detailed explanation.

[LuckyMD]

@aditisk that depends on what you put in adata.raw 😉. Initially adata.raw was used to store the full gene object when adata.X was filtered to only include HVGs or remove genes that aren’t expressed in enough cells. Now, we just have a boolean mask in adata.var['highly_variable'] for HVGs and so it’s often not used anymore. I typically store my log-normalized expression data there if I do batch correction or regress anything out, as adata.raw is used as default to compute rank_genes_groups and to show expression values on an embedding plot.