diff --git a/src/spac/visualization.py b/src/spac/visualization.py index b5984e30..04338b7a 100644 --- a/src/spac/visualization.py +++ b/src/spac/visualization.py @@ -400,7 +400,7 @@ def tsne_plot(adata, color_column=None, ax=None, **kwargs): def histogram(adata, feature=None, annotation=None, layer=None, group_by=None, together=False, ax=None, - x_log_scale=False, y_log_scale=False, **kwargs): + x_log_scale=False, y_log_scale=False, facet=False, **kwargs): """ Plot the histogram of cells based on a specific feature from adata.X or annotation from adata.obs. @@ -442,6 +442,9 @@ def histogram(adata, feature=None, annotation=None, layer=None, y_log_scale : bool, default False If True, the y-axis will be set to log scale. + facet : bool, default False + If True, group by function outputs facet plots + **kwargs Additional keyword arguments passed to seaborn histplot function. Key arguments include: @@ -562,7 +565,8 @@ def cal_bin_num( ): bins = max(int(2*(num_rows ** (1/3))), 1) print(f'Automatically calculated number of bins is: {bins}') - return(bins) + + return (bins) num_rows = plot_data.shape[0] @@ -624,6 +628,7 @@ def calculate_histogram(data, bins, bin_edges=None): if group_by: groups = df[group_by].dropna().unique().tolist() n_groups = len(groups) + if n_groups == 0: raise ValueError("There must be at least one group to create a" " histogram.") @@ -660,62 +665,86 @@ def calculate_histogram(data, bins, bin_edges=None): if feature: ax.set_title(f'Layer: {layer}') axs.append(ax) + else: - fig, ax_array = plt.subplots( - n_groups, 1, figsize=(5, 5 * n_groups) - ) + if not facet: + fig, ax_array = plt.subplots( + n_groups, 1, figsize=(5, 5 * n_groups) + ) + + # Convert a single Axes object to a list + # Ensure ax_array is always iterable + if n_groups == 1: + ax_array = [ax_array] - # Convert a single Axes object to a list - # Ensure ax_array is always iterable - if n_groups == 1: - ax_array = [ax_array] - else: - ax_array = ax_array.flatten() - - for i, ax_i in enumerate(ax_array): - group_data = plot_data[plot_data[group_by] == - groups[i]][data_column] - hist_data = calculate_histogram(group_data, kwargs['bins']) - - sns.histplot(data=hist_data, x="bin_center", ax=ax_i, - weights='count', **kwargs) - # If plotting feature specify which layer - if feature: - ax_i.set_title(f'{groups[i]} with Layer: {layer}') else: - ax_i.set_title(f'{groups[i]}') + ax_array = ax_array.flatten() + + for i, ax_i in enumerate(ax_array): + group_data = plot_data[plot_data[group_by] == + groups[i]][data_column] + hist_data = calculate_histogram(group_data, kwargs['bins']) + + sns.histplot(data=hist_data, x="bin_center", ax=ax_i, + weights='count', **kwargs) + # If plotting feature specify which layer + if feature: + ax_i.set_title(f'{groups[i]} with Layer: {layer}') + else: + ax_array = ax_array.flatten() + + # Set axis scales if y_log_scale is True + if y_log_scale: + ax_i.set_yscale('log') + + # Adjust x-axis label if x_log_scale is True + if x_log_scale: + xlabel = f'log({data_column})' + else: + xlabel = data_column + ax_i.set_xlabel(xlabel) + + # Adjust y-axis label based on 'stat' parameter + stat = kwargs.get('stat', 'count') + ylabel_map = { + 'count': 'Count', + 'frequency': 'Frequency', + 'density': 'Density', + 'probability': 'Probability' + } + ylabel = ylabel_map.get(stat, 'Count') + if y_log_scale: + ylabel = f'log({ylabel})' + ax_i.set_ylabel(ylabel) + axs.append(ax_i) + else: + hist = sns.FacetGrid(plot_data, col=group_by) + # Map the histogram function to the grid + hist.map(sns.histplot, data_column, **kwargs) - # Set axis scales if y_log_scale is True - if y_log_scale: - ax_i.set_yscale('log') + # Set rotation of label + hist.set_xticklabels(rotation=20, ha='right') + + # Titles for each facet + hist.set_titles("{col_name}") + + # Ajust top margin + hist.figure.subplots_adjust(left=.1, + top=0.85, + bottom=0.15, + hspace=0.3) + + fig = hist.figure + axs.extend(hist.axes.flat) + hist_data = plot_data - # Adjust x-axis label if x_log_scale is True - if x_log_scale: - xlabel = f'log({data_column})' - else: - xlabel = data_column - ax_i.set_xlabel(xlabel) - - # Adjust y-axis label based on 'stat' parameter - stat = kwargs.get('stat', 'count') - ylabel_map = { - 'count': 'Count', - 'frequency': 'Frequency', - 'density': 'Density', - 'probability': 'Probability' - } - ylabel = ylabel_map.get(stat, 'Count') - if y_log_scale: - ylabel = f'log({ylabel})' - ax_i.set_ylabel(ylabel) - - axs.append(ax_i) else: # Precompute histogram data for single plot hist_data = calculate_histogram(plot_data[data_column], kwargs['bins']) if pd.api.types.is_numeric_dtype(plot_data[data_column]): ax.set_xlim(hist_data['bin_left'].min(), - hist_data['bin_right'].max()) + hist_data['bin_right'].max()) + sns.histplot( data=hist_data, @@ -730,35 +759,38 @@ def calculate_histogram(data, bins, bin_edges=None): ax.set_title(f'Layer: {layer}') axs.append(ax) - # Set axis scales if y_log_scale is True - if y_log_scale: - ax.set_yscale('log') + axes = axs if isinstance(axs, (list, np.ndarray)) else [axs] + for ax in axes: + # Set axis scales if y_log_scale is True + if y_log_scale: + ax.set_yscale('log') - # Adjust x-axis label if x_log_scale is True - if x_log_scale: - xlabel = f'log({data_column})' - else: - xlabel = data_column - ax.set_xlabel(xlabel) - - # Adjust y-axis label based on 'stat' parameter - stat = kwargs.get('stat', 'count') - ylabel_map = { - 'count': 'Count', - 'frequency': 'Frequency', - 'density': 'Density', - 'probability': 'Probability' - } - ylabel = ylabel_map.get(stat, 'Count') - if y_log_scale: - ylabel = f'log({ylabel})' - ax.set_ylabel(ylabel) + # Adjust x-axis label if x_log_scale is True + if x_log_scale: + xlabel = f'log({data_column})' + else: + xlabel = data_column + ax.set_xlabel(xlabel) + + # Adjust y-axis label based on 'stat' parameter + stat = kwargs.get('stat', 'count') + ylabel_map = { + 'count': 'Count', + 'frequency': 'Frequency', + 'density': 'Density', + 'probability': 'Probability' + } + ylabel = ylabel_map.get(stat, 'Count') + if y_log_scale: + ylabel = f'log({ylabel})' + ax.set_ylabel(ylabel) if len(axs) == 1: return {"fig": fig, "axs": axs[0], "df": hist_data} else: return {"fig": fig, "axs": axs, "df": hist_data} + def heatmap(adata, column, layer=None, **kwargs): """ Plot the heatmap of the mean feature of cells that belong to a `column`. diff --git a/tests/test_visualization/test_histogram.py b/tests/test_visualization/test_histogram.py index f8ba95ea..999b66e5 100644 --- a/tests/test_visualization/test_histogram.py +++ b/tests/test_visualization/test_histogram.py @@ -222,7 +222,7 @@ def test_y_log_scale_axis(self): def test_y_log_scale_label(self): """Test that y-axis label is updated when y_log_scale is True.""" - fig, ax, dfd = histogram( + fig, ax, df = histogram( self.adata, feature='marker1', y_log_scale=True @@ -413,6 +413,44 @@ def test_default_bins_calculation(self): expected_bins = max(int(2 * (self.adata.shape[0] ** (1 / 3))), 1) self.assertEqual(n_bins, expected_bins) + def test_facet_plot(self): + """Test that facet plot works.""" + fig, ax, df = histogram( + self.adata, + feature='marker1', + group_by='annotation2', + facet=True, + ).values() + + # Check if axs is a collection (list/array of Axes) + self.assertIsInstance(ax, (list, np.ndarray), + "Output is not a multi-axis grid") + + # Check number of facets equals number of unique groups + unique_groups = self.adata.obs['annotation2'].dropna().unique() + self.assertEqual(len(ax), len(unique_groups), + f"Expected {len(unique_groups)}" + f" facet plots, got {len(ax)}.") + + # Validate each axis: title, xlabel, and ylabel + for i, axis in enumerate(ax): + # Check that title is set and matches the group + title = axis.get_title() + self.assertTrue(title, f"Facet {i} is missing a title.") + self.assertTrue(any(str(group) in title + for group in unique_groups), + f"Title '{title}' does not contain" + f"any expected group names.") + + # Check X and Y labels + self.assertIn('marker1', axis.get_xlabel(), + f"Facet {i} X-axis label" + f" '{axis.get_xlabel()}' is incorrect.") + self.assertIn(axis.get_ylabel(), + ['Count', 'Frequency', 'Density', 'Probability'], + f"Facet {i} Y-axis label" + f" '{axis.get_ylabel()}' is not a valid stat.") + if __name__ == '__main__': unittest.main()