crested.pl.qc.filter_cutoff

crested.pl.qc.filter_cutoff(adata, cutoffs=(1.5, 1, 0.5, 0), model_name=None, color_points=True, double_y_axis=True, line_cmap='tab10', plot_kws=None, line_kws=None, ax=None, **kwargs)

Plot the effect of different potential filtering cutoffs in filter_regions_on_specificity() before doing the filtering.

If the default cutoff (1 standard deviation above the mean) is too stringent or too lenient, try different cutoffs and pick the best one.

Parameters:

• adata (AnnData) – AnnData object with region data. This should not be filtered yet!

• cutoffs (Sequence (default: (1.5, 1, 0.5, 0))) – List of considered gini standard deviation cutoffs to plot, as in filter_regions_on_specificity()’s gini_std_threshold (where the default is 1).

• model_name (str | None (default: None)) – The name of the model to calculate scores from. If None or ‘truth’/’groundtruth’/’X’ (default), will use the values in adata.X.

• color_points (bool (default: True)) – Whether to color the region points according to their position inside/outside the different cutoffs. If True (default), adds 'c': cutoff_index, 'cmap': 'Greys' and 'vmin': 0 to plot_kws (unless manually specified).

• double_y_axis (bool (default: True)) – Whether to add a secondary y axis on the right side showing the Gini scores, augmenting the default axis showing standard deviations from the Gini mean.

• line_cmap (str | Sequence (default: 'tab10')) – Colors or colormap to draw colors for the different cutoffs lines from. Any valid input to color_palette() works. Examples are a matplotlib or seaborn colormap name or a list of colors.

• plot_kws (dict | None (default: None)) – Extra keyword arguments passed to scatter(). Defaults: 's': 4.

• line_kws (dict | None (default: None)) – Extra keyword arguments passed to axvline(). Defaults: 'linestyle': '--'.

• ax (plt.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 8.

• height – Height of the newly created figure if ax=None. Default is 6.

• kwargs – Additional arguments passed to render_plot() to control the final plot output. Please see render_plot() for details. Custom defaults for gini_filtering: xlabel='Rank', ylabel='Standard deviations from the mean Gini', grid='x'.

Return type:

(plt.Figure, plt.Axes) | None

See also

crested.pl.render_plot, crested.pp.filter_regions_on_specificity

Example

>>> crested.pl.qc.filter_cutoff(adata, cutoffs=[2, 1.5, 1])