Added more examples in the benchmark and new notebooks calling mofa+ …

…and gfa.

examples/ism_utils.py

@@ -0,0 +1,67 @@
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+
+def specificity_score(df_w):
+    n_themes = (df_w.shape[1]-1)
+    k = np.unique(df_w['list_cell_codes']).shape[0]
+    means_within = np.zeros(n_themes)
+    means_outside = np.zeros(n_themes)
+    specificity = np.zeros((k,n_themes))
+    group_means = df_w.iloc[:, 0:n_themes+1].groupby('list_cell_codes').mean()
+    for i in range(k):
+        for j in range(n_themes):
+            means_within[j] = group_means.iloc[i,j]
+            means_outside[j] = np.mean(np.hstack((group_means.iloc[:i-1,j],group_means.iloc[i+1:,j])))
+
+        specificity[i,:] = (means_within - means_outside) / np.max(means_within)
+        # specificity[specificity<0] = 0
+
+        means_within_outside = np.hstack((means_within[:,np.newaxis],
+                                      means_outside[:,np.newaxis], specificity[i,:][:,np.newaxis]))
+
+        if i==0:
+            df_means_within_outside = pd.DataFrame(means_within_outside.copy())
+            df_means_within_outside.insert(0, 'cell_type', i)
+            df_means_within_outside.insert(0, 'id', range(n_themes))
+        else:
+            df_means_within_outside_2 = pd.DataFrame(means_within_outside.copy())
+            df_means_within_outside_2.insert(0, 'cell_type', i)
+            df_means_within_outside_2.insert(0, 'id', range(n_themes))
+            df_means_within_outside = pd.concat([df_means_within_outside, df_means_within_outside_2], ignore_index=True)
+
+    df_means_within_outside.columns = ['id', 'cell_type', 'within', 'outside', 'specificity']
+    df_means_within_outside['cell_type'] = df_means_within_outside['cell_type'].astype('category')
+    df_means_within_outside['id'] = df_means_within_outside['id'].astype('category')
+
+    return specificity, df_means_within_outside
+
+def specificity_plot(df_means_within_outside):
+    x = df_means_within_outside['outside']
+    y = df_means_within_outside['within']
+    z = df_means_within_outside['cell_type']
+    ids = df_means_within_outside['id']
+
+    # Create subplots for each category of z
+    unique_categories = np.unique(z)
+    num_categories = len(unique_categories)
+
+    fig, axs = plt.subplots(1, num_categories, figsize=(14, 2), sharex=False, sharey=False)
+
+    for i, category in enumerate(unique_categories):
+        mask = (z == category)
+        min_val = .95*min(x[mask].min(), y[mask].min())
+        min_val = 0
+        max_val = 1.05*max(x[mask].max(), y[mask].max())
+        axs[i].scatter(x[mask], y[mask], label=f'Category {category}')
+        axs[i].plot([min_val, max_val], [min_val, max_val], ls='--', c='r')  # Add the diagonal line (identity line)
+        # for ii, row in enumerate(ids[mask]):
+        #     axs[i].text(x[ii], y[ii], row, fontsize=8)
+        axs[i].set_xlim(min_val, max_val)
+        axs[i].set_ylim(min_val, max_val)
+        axs[i].set_title(f'Cell type ({category})')
+        axs[i].set_xlabel('Outside')
+        axs[i].set_ylabel('Within')
+
+    plt.tight_layout()
+    plt.show()