Facecolor of scatter plots for categorical data fail to be non-transparent.

Hi,

Even when we set sc.settings.set_figure_params(transparent=False) the scatter plot for CATEGORYCAL data is transparented. (ex. sc.pl.umap(adata, color='louvain') ) Continuous data goes well, so I think it is a bug.

When you switch Jupyterlab’s theme into dark, it will be easy to check.

Best, Yoshiaki

[flying-sheep]

Expliciting that ‘white’, ‘w’ or (1,1,1) are also applicable may be kind.

Definitely! Good docs are everything.

[yyoshiaki]

I was wrong, it worked well…

%matplotlib inline

import scanpy as sc
import matplotlib as mpl

# 2 lines below solved the facecolor problem.
mpl.rcParams['figure.facecolor'] = 'white'
sc.settings.set_figure_params(dpi=80, color_map='viridis', transparent=False)

adata = sc.datasets.paul15()
sc.pp.recipe_zheng17(adata)
sc.tl.pca(adata, svd_solver='arpack')
sc.pl.pca(adata, color='paul15_clusters', legend_loc='on data')

Anyway,

So we should add a facecolor parameter and deprecate transparent (with a nice message to point people to set_figure_params(facecolor=(0, 0, 0, 0)))

sounds a good solution. Expliciting that ‘white’, ‘w’ or (1,1,1) are also applicable may be kind. Of course, you can add theme like Seurat’s one though I’m not sure how many people are requiring it.

• https://satijalab.org/seurat/mca.html
• https://matplotlib.org/gallery/style_sheets/style_sheets_reference.html#sphx-glr-gallery-style-sheets-style-sheets-reference-py