cell_topic_heatmap color_dictionary can't be None

[camiel-m]

129 I ran into this same issue and added some code to check if the color_dict exists and if not to generate a simple one using matplotlib.colormaps

[SeppeDeWinter]

Thanks for the PR!

Looking again at the code I noticed that the case where variables is None is also not handled correctly (hence the extra changes).

When a variable is not present in the color_dictionary a random color was already assigned. If you prefer this to be colors from tab20, or another color map from matplotlib, feel free to change this.

@camiel-m could you check wether the code is working?

Thanks,

Seppe

[camiel-m]

color_dict = None is now working fine, but if you don't specify any variables you get the following error:

---------------------------------------------------------------------------

KeyError                                  Traceback (most recent call last)
Cell In[13], line 2
      1 key = 'pycisTopic_leiden_10_0.3'
----> 2 cell_topic_heatmap(
      3     cistopic_obj,
      4     scale = False,
      5     legend_loc_x = 1.0,
      6     legend_loc_y = -1.2,
      7     legend_dist_y = -1,
      8     figsize = (10, 10)
      9 )

File #/git_repos/pycisTopic/src/pycisTopic/clust_vis.py:933, in cell_topic_heatmap(cistopic_obj, variables, remove_nan, scale, cluster_topics, color_dictionary, seed, legend_loc_x, legend_loc_y, legend_dist_y, figsize, selected_topics, selected_cells, harmony, save)
    926 if scale:
    927     cell_topic = pd.DataFrame(
    928         sklearn.preprocessing.StandardScaler().fit_transform(cell_topic),
    929         index=cell_topic.index.to_list(),
    930         columns=cell_topic.columns,
    931     )
--> 933 if remove_nan and (sum(cell_data[variables].isnull().sum()) > 0):
    934     cell_data = cell_data[variables].dropna()
    935     cell_topic = cell_topic.loc[:, cell_data.index.tolist()]

maybe a simple change like this would be enough?

    if variables:
        if remove_nan and (sum(cell_data[variables].isnull().sum()) > 0):
            cell_data = cell_data[variables].dropna()
            cell_topic = cell_topic.loc[:, cell_data.index.tolist()]

[SeppeDeWinter]

Ah yes, right!

Should be fixed now (I hope).

[camiel-m]

It doesn't throw an error, but the plot does look quite different from default behavior. The axis are flipped and the cells are not sorted. Is this the intended behavior or should the cells maybe sorted based on similarity or by highest topic score? Then again I don't think a lot of people will use this without cluster labels.

[SeppeDeWinter]

hmm yes. does not look super informative.

How about when you set cluster_topics to True (it's False by default)

[camiel-m]

It's a lot more interpretable that way.

[SeppeDeWinter]

Great!!