test if not none

Author: dbdimitrov

src/liana/method/_pipe_utils/_get_mean_perms.py

@@ -64,20 +64,20 @@ def _permute_and_aggregate(perm, perm_idx, X, labels_mask, agg_fun):
 
 
 def _generate_perms_cube(X, n_perms, labels_mask, seed, agg_fun, n_jobs, verbose):
+    # initialize rng
+    rng = np.random.default_rng(seed=seed)
+
     # indexes to be shuffled
     idx = np.arange(X.shape[0])
+
     # Perm should be a cube /w dims: n_perms x idents x n_genes
     perms = np.zeros((n_perms, labels_mask.shape[1], X.shape[1]))
 
     # Use Parallel to enable parallelization
-    if n_jobs == 1:
-        results = [_permute_and_aggregate(perm, np.random.default_rng(seed=seed + perm).permutation(idx), X, labels_mask, agg_fun)
-                   for perm in tqdm(range(n_perms), disable=not verbose)]
-    else:
-        results = Parallel(n_jobs=n_jobs)(delayed(_permute_and_aggregate)
-                                          (perm, np.random.default_rng(seed=seed + perm).permutation(idx), X, labels_mask, agg_fun)
-                                          for perm in tqdm(range(n_perms), disable=not verbose)
-                                          )
+    results = Parallel(n_jobs=n_jobs)(delayed(_permute_and_aggregate)
+                                      (perm, rng.permutation(idx), X, labels_mask, agg_fun)
+                                      for perm in tqdm(range(n_perms), disable=not verbose)
+                                      )
 
     # Unpack results
     for perm, permuted_means in results:
@@ -135,7 +135,6 @@ def _calculate_pvals(lr_truth, perm_stats, _score_fun):
         n_perms = perm_stats.shape[1]
         pvals = np.sum(np.greater_equal(lr_perm_means, lr_truth), axis=0) / n_perms
     else:
-        pvals = None
-
+        pvals = np.ones(lr_truth.shape)
 
     return pvals