Plotting colormap not compatible with scanpy

[casblaauw]

Report

Following #8 , tf-mindi saves the colormap to adata.uns[f'{your_variable}_colors'], in an attempt to align with scanpy. I think saving the colormap makes a ton of sense, except that sadly scanpy for a mysterious reason has decided to only save the list of colors, and regenerate the associated keys with pd.Categorical(adata.obs['your_variable']).categories every time, just assuming that they still match.

Trying to use the scanpy plotting function now throws a warning and overwrites your colors:

> sc.pl.tsne(adata, color='cluster_dbd')
WARNING: The following color value found in adata.uns['cluster_dbd_colors'] is not valid: 'C2H2 ZF'. Default colors will be used instead.

I can have a look at fixing this in ensure_colors (maintaining compatibility with existing adatas containing the dict colormap, of course), but it'll have to be in a few weeks. For now, sc.pl.tsne(adata, color='cluster_dbd', palette=adata.uns['cluster_dbd_colors']) should work.

Versions

[casblaauw]

Okay if you use palette=adata.uns['cluster_dbd_colors'] with the TF-MINDI generated colormap, it overwrites it with the list-version from scanpy, and then you can't plot in TF-MINDI anymore because it expects a dict:

> sc.pl.tsne(adata, color='cluster_dbd',  palette=adata.uns['cluster_dbd_colors'])
> tm.pl.tsne(adata, color_by="cluster_dbd")
File /lustre1/project/stg_00002/lcb/cblaauw/python_modules/TF-MINDI/src/tfmindi/pl/tsne.py:210, in tsne_logos(adata, patterns, color_by, logo_width, logo_height, alpha, s, ic_threshold, min_nucleotides, min_seqlets, show_cluster_labels, show_legend, gray_background, cmap, **kwargs)
    207 else:
    208     # Use stored color management, fallback to leiden if color_by column doesn't exist
    209     try:
--> 210         point_colors, color_map = get_point_colors(adata, color_by, cmap)
    211         color_column = color_by
    212     except ValueError:
    213         # Fallback to leiden clustering colors if color_by column doesn't exist

File /lustre1/project/stg_00002/lcb/cblaauw/python_modules/TF-MINDI/src/tfmindi/pl/_utils.py:379, in get_point_colors(adata, column, cmap, use_stored_colors)
    377 if color_values.dtype == "category" or color_values.dtype == object:
    378     if use_stored_colors:
--> 379         color_map = get_colors(adata, column, cmap)
    380     else:
    381         # Generate fresh colors without storing
    382         if color_values.dtype == "category":
File /lustre1/project/stg_00002/lcb/cblaauw/python_modules/TF-MINDI/src/tfmindi/pl/_utils.py:286, in get_colors(adata, column, cmap)
    264 def get_colors(
    265     adata: AnnData,
    266     column: str,
    267     cmap: str = "tab20",
    268 ) -> dict[str, str]:
    269     """
    270     Get colors for a categorical column, generating them if they don't exist.
    271 
   (...)    284         Dictionary mapping category values to hex color codes
    285     """
--> 286     return ensure_colors(adata, column, cmap, force_regenerate=False)

File /lustre1/project/stg_00002/lcb/cblaauw/python_modules/TF-MINDI/src/tfmindi/pl/_utils.py:203, in ensure_colors(adata, column, cmap, force_regenerate)
    200     if values.isnull().any():
    201         current_categories.add("Unknown")
--> 203 existing_categories = set(existing_colors.keys())
    205 # If all current categories have colors, return existing colormap
    206 if current_categories.issubset(existing_categories):
AttributeError: 'list' object has no attribute 'keys'