Update to 0.2.0:

add circular fingerprints.

mgktools/__init__.py

@@ -2,4 +2,4 @@
 # -*- coding: utf-8 -*-
 
 
-__version__ = '0.1.1'
+__version__ = '0.2.0'

mgktools/features_mol/features_generators.py

@@ -5,6 +5,7 @@
 import numpy as np
 from rdkit import Chem, DataStructs
 from rdkit.Chem import AllChem, Descriptors
+import deepchem
 from descriptastorus.descriptors import rdDescriptors, rdNormalizedDescriptors
 
 
@@ -15,7 +16,7 @@ def __init__(self, features_generator_name: Union[str, Callable],
         self.features_generator_name = features_generator_name
         self.radius = radius
         self.num_bits = num_bits
-        if features_generator_name in ['morgan', 'morgan_count']:
+        if features_generator_name in ['morgan', 'morgan_count', 'circular']:
             assert self.radius is not None
             assert self.num_bits is not None
 
@@ -26,6 +27,8 @@ def __call__(self, mol: Union[str, Chem.Mol]) -> np.ndarray:
             return self.morgan_binary_features_generator(mol)
         elif self.features_generator_name == 'morgan_count':
             return self.morgan_counts_features_generator(mol)
+        elif self.features_generator_name == 'circular':
+            return self.circular_features_generator(mol)
         elif self.features_generator_name == 'rdkit_208':
             return self.rdkit_208_features_generator(mol)
         elif self.features_generator_name == 'rdkit_2d':
@@ -67,6 +70,13 @@ def morgan_counts_features_generator(self, mol: Union[str, Chem.Mol]) -> np.ndar
 
         return features
 
+    @staticmethod
+    def circular_features_generator(mol: Union[str, Chem.Mol]) -> np.ndarray:
+        circular_fp_featurizer = deepchem.feat.CircularFingerprint(size=2048, radius=8,
+                                                                   sparse=False, smiles=True)
+        features = circular_fp_featurizer.featurize([mol]).ravel()
+        return features
+
     @staticmethod
     def rdkit_2d_features_generator(mol: Union[str, Chem.Mol]) -> np.ndarray:
         """

setup.py

@@ -32,6 +32,7 @@ def read(*filenames, **kwargs):
         'rxntools>=0.0.2',
         'pycuda>=2022.1',
         'rdkit>=2022.9.2',
+        'deepchem==2.7.2.dev20231207083329'
     ],
     author='Yan Xiang',
     author_email='1993.xiangyan@gmail.com',

test/cross_validation/test_cv_pure.py

@@ -11,18 +11,19 @@
 from mgktools.evaluators.cross_validation import Evaluator
 
 
-pure = ['CCCC', 'CCCCCO', 'c1ccccc1', 'CCNCCO', 'CCCCN', 'NCCCCCO', 'c1ccccc1N', 'NCCNCCO']
-targets_regression = [3.1, 14.5, 25.6, 56.7, 9.1, 17.5, 22.6, 36.7]
+pure = ['CCCC', 'CCCCCO', 'c1ccccc1', 'CCNCCO', 'CCCCN', 'NCCCCCO', 'c1ccccc1N', 'NCCNCCO',
+        'CNC(CC)CC', 'c1ccccc1', 'c1ccccc1CCCCc1ccccc1', 'CC(=O)OCCO']
+targets_regression = [3.1, 14.5, 25.6, 56.7, 9.1, 17.5, 22.6, 36.7, 23.1, 32.1, 1.4, 7.6]
 df_regression = pd.DataFrame({'pure': pure, 'targets': targets_regression})
-targets_classification = [1, 1, 0, 1, 1, 0, 0, 1]
+targets_classification = [1, 1, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1]
 df_classification = pd.DataFrame({'pure': pure, 'targets': targets_classification})
 
 
 @pytest.mark.parametrize('mgk_file', [additive_norm, additive_pnorm, additive_msnorm,
                                       product_norm, product_pnorm, product_msnorm])
 @pytest.mark.parametrize('model', ['gpc', 'svc'])
 @pytest.mark.parametrize('split_type', ['random', 'scaffold_order', 'scaffold_random'])
-def test_only_graph(mgk_file, model, split_type):
+def test_only_graph_classification(mgk_file, model, split_type):
     dataset = Dataset.from_df(df=df_classification,
                               pure_columns=['pure'],
                               target_columns=['targets'])
@@ -49,12 +50,12 @@ def test_only_graph(mgk_file, model, split_type):
                                       product_norm, product_pnorm, product_msnorm])
 @pytest.mark.parametrize('modelsets', [('gpr', None, None, None),
                                        ('gpr-sod', 2, 3, 'smallest_uncertainty'),
-                                       ('gpr-sod', 2, 3, 'weight_uncertainty'),
+                                       # ('gpr-sod', 2, 3, 'weight_uncertainty'),
                                        ('gpr-sod', 2, 3, 'mean'),
                                        ('gpr-nystrom', None, 3, None),
                                        ('gpr-nle', None, 3, None)])
 @pytest.mark.parametrize('split_type', ['random', 'scaffold_order', 'scaffold_random'])
-def test_only_graph(mgk_file, modelsets, split_type):
+def test_only_graph_scalable_gps(mgk_file, modelsets, split_type):
     model_type, n_estimators, n_samples, consensus_rule = modelsets
     dataset = Dataset.from_df(df=df_regression,
                               pure_columns=['pure'],

test/data/test_data_pure.py

@@ -27,6 +27,7 @@ def test_only_graph(testset):
 
 @pytest.mark.parametrize('testset', [
     ('morgan', 2048),
+    ('circular', 2048),
     ('rdkit_2d', 200),
     ('rdkit_2d_normalized', 200),
 ])
@@ -42,6 +43,7 @@ def test_only_fingerprints(testset):
 
 @pytest.mark.parametrize('testset', [
     ('morgan', 2048),
+    ('circular', 2048),
     ('rdkit_2d', 200),
     ('rdkit_2d_normalized', 200),
 ])