Add morgan fingerprints

assets/evaluation_metrics.json

@@ -1,7 +1,7 @@
 {
-    "Mean Squared Error": 0.462,
-    "Root Mean Squared Error": 0.6797,
-    "Mean Absolute Error": 0.4891,
-    "R-squared Score": 0.6514,
-    "Explained Variance Score": 0.6521
+    "Mean Squared Error": 0.3952,
+    "Root Mean Squared Error": 0.6287,
+    "Mean Absolute Error": 0.4569,
+    "R-squared Score": 0.7018,
+    "Explained Variance Score": 0.7018
 }

scripts/testing.py

@@ -10,6 +10,7 @@
 from rdkitclassical_desc import RDkitClassicalDescriptor
 from mordred_desc import MordredDescriptor
 from regressor import Regressor
+from morgan_desc import MorganFingerprint
 
 output_folder = os.path.join(root, "..", "assets")
 
@@ -31,7 +32,7 @@
     y_valid.reset_index(drop=True, inplace=True)
 
     # Instantiate the descriptor class
-    descriptor = RDkitClassicalDescriptor()
+    descriptor = MorganFingerprint()
 
     descriptor.fit(smiles_train)
 
@@ -40,7 +41,7 @@
     smiles_valid_transformed = descriptor.transform(smiles_valid)
 
     # Instantiate the regressor
-    regressor = Regressor(output_folder, algorithm='xgboost', k=52)
+    regressor = Regressor(output_folder, algorithm='catboost')
 
     # Train the model 
     regressor.fit(smiles_train_transformed, y_train) 

xai4chem/morgan_desc.py

@@ -0,0 +1,52 @@
+import numpy as np
+import pandas as pd
+from rdkit.Chem import rdFingerprintGenerator
+from sklearn.feature_selection import VarianceThreshold
+
+from rdkit.Chem import rdMolDescriptors as rd
+from rdkit import Chem
+
+RADIUS = 3
+NBITS = 2048
+DTYPE = np.uint8
+
+def clip_sparse(vect, nbits):
+    l = [0]*nbits
+    for i,v in vect.GetNonzeroElements().items():
+        l[i] = v if v < 255 else 255
+    return l
+
+
+class _Fingerprinter(object):
+
+    def __init__(self):
+        self.nbits = NBITS
+        self.radius = RADIUS
+
+    def calc(self, mol):
+        v = rd.GetHashedMorganFingerprint(mol, radius=self.radius, nBits=self.nbits)
+        return clip_sparse(v, self.nbits)
+
+
+def morgan_featurizer(smiles):
+    d = _Fingerprinter()
+    X = np.zeros((len(smiles), NBITS))
+    for i, smi in enumerate(smiles):
+        mol = Chem.MolFromSmiles(smi)
+        X[i,:] = d.calc(mol)
+    return X
+
+
+class MorganFingerprint(object):
+
+    def __init__(self):
+        pass
+
+    def fit(self, smiles):
+        X = morgan_featurizer(smiles)
+        self.features = ["fp-{0}".format(i+1) for i in range(X.shape[1])]
+        return pd.DataFrame(X, columns=self.features)
+
+    def transform(self, smiles):
+        X = morgan_featurizer(smiles)
+        return pd.DataFrame(X, columns=self.features)

xai4chem/regressor.py

@@ -101,7 +101,7 @@ def _train_and_evaluate_optuna_model(self, model, X, y):
 
     def fit(self, X_train, y_train, default_params=True):
         self._select_features(X_train, y_train)
-        print(f'Selected features: {self.selected_features}')
+        # print(f'Selected features: {self.selected_features}')
         X_train = X_train[self.selected_features]
 
         if self.algorithm == 'xgboost':