crested.pl.corr.scatter

Contents

crested.pl.corr.scatter#

crested.pl.corr.scatter(adata, class_name=None, model_names=None, split='test', log_transform=False, exclude_zeros=True, density_indication=False, square=False, identity_line=False, cbar=False, downsample_density=10000, max_threads=8, plot_kws=None, cbar_kws=None, ax=None, **kwargs)#

Plot a density scatter plot of predictions vs ground truth for specified models and class.

Parameters:

  • adata (AnnData) – AnnData object containing the data in X and predictions in layers.

  • class_name (str | None (default: None)) – Name of the class in adata.obs_names. If None, plot is made for all the classes.

  • model_names (str | list[str] | None (default: None)) – Model name or list of model names in adata.layers. If None, will create a plot per model in adata.layers.

  • split (str | None (default: 'test')) – ‘train’, ‘val’, ‘test’ subset or None. If None, will use all splits. If not None, expects a “split” column in adata.var.

  • log_transform (bool (default: False)) – Whether to log-transform the data before plotting.

  • exclude_zeros (bool (default: True)) – Whether to exclude zero ground truth values from the plot.

  • density_indication (bool (default: False)) – Whether to indicate density in the scatter plot.

  • square (bool (default: False)) – Whether to force the plots to be square, have equal aspect ratios, and equal shared axis ranges.

  • identity_line (bool (default: False)) – Whether to plot a y=x line denoting perfect correlation.

  • cbar (bool (default: False)) – Whether to plot the colorbar when using density_indication.

  • downsample_density (int (default: 10000)) – Number of points to downsample to when fitting the density if using the density indication. Note that one point denotes one region for one class, so the full set would be # of (test) regions * # classes. Default is 10000. If False, will not downsample.

  • max_threads (int (default: 8)) – Maximum number of threads to use when evaluating the density if using the density indication. If 1, will not parallelize.

  • plot_kws (dict | None (default: None)) – Extra keyword arguments passed to scatter(). Defaults: {'alpha': 0.25, 'edgecolor': 'k'}.

  • cbar_kws (dict | None (default: None)) – Extra keyworde arguments passed to colorbar(). Defaults: {'label': 'Density', 'shrink': 0.8}

  • ax (Axes | None (default: None)) – Axis to plot values on. If not supplied, creates a figure from scratch.

  • width – Width of the newly created figure if ax=None. Default is 7 per model without cbar, or 8 with cbar.

  • height – Height of the newly created figure if ax=None. Default is 8.

  • sharex – Whether to share the x axes of the created plots. Default is False. Setting square=True does equalize limits even if sharex=False

  • sharey – Whether to share the y axes of the created plots. Default is True. Setting square=True does equalize limits even if sharey=False

  • kwargs – Additional arguments passed to render_plot() to control the final plot output. Please see render_plot() for details. Custom defaults for class_density: xlabel="Ground truth", ylabel='Predictions', alpha='0.25', title=({class_name} - ){model_name}, suptitle='Targets vs predictions (for {class_name})' (if n_models>1).

Return type:

tuple[Figure, Axes] | tuple[Figure, list[Axes]] | None

See also

crested.pl.render_plot

Example

>>> crested.pl.corr.scatter(
...     adata,
...     class_name="Astro",
...     model_names=["Base model", "Fine-tuned"],
...     split="test",
...     log_transform=True,
...     density_indication=True,
...     square=True,
...     identity_line=True,
...     save_path="temp_figs/corr_scatter.png"
... )
../../_images/corr_scatter.png
Contents