add random_seed to permute_data (#221)

CHANGELOG.md

@@ -13,6 +13,7 @@ Incompatible changes:
 * Other backwards incompatible function API changes and code refactoring.
 
 New features:
+* Add random_seed option to tranforming.permute_data()
 * Add bad_color parameter to heatmap() and derivative functions
 * Add more methods for MS1Experiment
 * Add q-values (correted p-values) to dsfdr and derivative functions. This is manifested in a new feature_metadata field ("qval") for results of diff_abundance() / correlation()

calour/transforming.py

@@ -249,7 +249,7 @@ def log_n(exp: Experiment, n=1, inplace=False) -> Experiment:
     return exp
 
 
-def permute_data(exp: Experiment, normalize=True, inplace=False) -> Experiment:
+def permute_data(exp: Experiment, normalize=True, inplace=False, random_seed=None) -> Experiment:
     '''Shuffle independently the abundances of each feature.
 
     This creates a new experiment with no dependency between features.
@@ -259,19 +259,27 @@ def permute_data(exp: Experiment, normalize=True, inplace=False) -> Experiment:
     normalize : bool, optional
         True (default) to normalize each sample after completing the feature shuffling.
         False to not normalize
+    random_seed : int, np.radnom.Generator instance or None, optional, default=None
+        set the random number generator seed for the random permutations
+        If int, random_seed is the seed used by the random number generator;
+        If Generator instance, random_seed is set to the random number generator;
+        If None, then fresh, unpredictable entropy will be pulled from the OS
 
     Returns
     -------
     Experiment
         With each feature shuffled independently
 
     '''
+    # create the numpy.random.Generator
+    rng = np.random.default_rng(random_seed)
+
     if not inplace:
         exp = deepcopy(exp)
 
     exp.sparse = False
     for cfeature in range(exp.shape[1]):
-        np.random.shuffle(exp.data[:, cfeature])
+        rng.shuffle(exp.data[:, cfeature])
     if normalize:
         exp.normalize(np.mean(exp.data.sum(axis=1)), inplace=True)
     return exp