Merge branch 'small_bug_fixes' into 'main'

Small bug fixes

See merge request cdd/QSPRpred!93

CHANGELOG.md

@@ -12,6 +12,7 @@ From v1.3.1 to v2.0.0
 - fixed serialization issues with `DataFrameDescriptorSet` and saving and loading of MSA for PCM descriptor calculations 
 - the Papyrus adapter was fixed so that the quality and data set filtering options work properly (before only high quality Papyrus++ data was fetched no matter the options)
 - previously, in some cases cross-validation splits might not have been shuffled during hyperparameter optimization and evaluation on cross-validation folds (this might have resulted in suboptimal cross-validation performance and bad choices of hyperparameters), a fix was made in b029e78009d1fa7fdc694e388f244eb0ee1d8cc0
+- score_func can now be set in `QSPRModel`.
 
 ## Changes
 
@@ -38,6 +39,7 @@ From v1.3.1 to v2.0.0
   have a variance of 0.1 will be removed.
 - Added the ExtendedValenceSignature molecular descriptor based on Jean-Loup Faulon's work.
 - removed default parameter setting scikit-learn SVC and SVR `max_iter` 10000.
+- added `matthews_corrcoef` to the supported metrics for binary classification.
 
 ## New Features
 - New feature split `ManualSplit` for splitting data by a user-defined column

qsprpred/data/utils/descriptor_utils/fingerprints.py

@@ -3,7 +3,7 @@
 import numpy as np
 from rdkit import DataStructs
 from rdkit.Avalon import pyAvalonTools
-from rdkit.Chem import AllChem, rdMolDescriptors, rdmolops
+from rdkit.Chem import AllChem, MACCSkeys, rdMolDescriptors, rdmolops
 
 from .interfaces import Fingerprint
 
@@ -45,6 +45,38 @@ def __len__(self):
     def getKey(self):
         return "MorganFP"
 
+class RDKitMACCSFP(Fingerprint):
+    """RDKits implementation of MACCS keys fingerprint."""
+
+    def getFingerprints(self, mols):
+        """Return the MACCS fingerprints for the input molecules.
+
+        Args:
+            mols: molecules to obtain the fingerprint of
+
+        Returns:
+            fingerprint (list): `list` of fingerprints for "mols"
+        """
+        convertFP = DataStructs.ConvertToNumpyArray
+
+        ret = np.zeros((len(mols), len(self)))
+        for idx, mol in enumerate(mols):
+            fp = MACCSkeys.GenMACCSKeys(mol)
+            np_fp = np.zeros(len(fp))
+            convertFP(fp, np_fp)
+            ret[idx] = np_fp
+
+        return ret
+
+    @property
+    def settings(self):
+        return {}
+
+    def __len__(self):
+        return 167
+
+    def getKey(self):
+        return "RDKitMACCSFP"
 
 class MaccsFP(Fingerprint):
     def __init__(self, nBits=167, **kwargs):
@@ -267,7 +299,6 @@ def getFingerprint(self, fp_type, *args, **kwargs):
 
     def getMorganFP(self, *args, **kwargs):
         return MorganFP(*args, **kwargs)
-
     def getMaccsFP(self, *args, **kwargs):
         return MaccsFP(*args, **kwargs)
 
@@ -288,6 +319,9 @@ def getPatternFP(self, *args, **kwargs):
 
     def getLayeredFP(self, *args, **kwargs):
         return LayeredFP(*args, **kwargs)
+
+    def getRDKitMACCSFP(self, *args, **kwargs):
+        return RDKitMACCSFP(*args, **kwargs)
 
     def getCDKFP(self, *args, **kwargs):
         from qsprpred.extra.data.utils.descriptor_utils.fingerprints import CDKFP

qsprpred/deep/models/models.py

@@ -272,8 +272,9 @@ def evaluate(
         save: bool = True,
         es_val_size: float = 0.1,
         parameters: dict | None = None,
+        score_func=None
     ) -> np.ndarray:
-        """Make predictions for cross-validation and independent test set.
+        """Make predictions for crossvalidation and independent test set.
 
         Args:
             save (bool):
@@ -286,8 +287,12 @@ def evaluate(
         Returns:
             np.ndarray:
                 predictions for test set and cross-validation for further analysis
+            score_func (Metric):
+                scoring function for the model, if None, the default scoring function
+                for the task is used
         """
         evalparams = self.parameters if parameters is None else parameters
+        score_func = self.scoreFunc if score_func is None else score_func
         X, X_ind = self.data.getFeatures()
         y, y_ind = self.data.getTargetPropertiesValues()
         last_save_epochs = 0

qsprpred/model_CLI.py

@@ -298,7 +298,8 @@ def QSPR_modelling(args):
                 counts = mydataset.y.value_counts()
                 scale_pos_weight = (
                     counts[0] / counts[1] if
-                    (args.sample_weighing and not mydataset.isMultiClass()) else 1
+                    (args.sample_weighing and len(tasks)==1 and
+                     not tasks[0].isMultiClass()) else 1
                 )
                 if alg_dict[model_type] == XGBClassifier:
                     parameters["scale_pos_weight"] = scale_pos_weight

qsprpred/models/hyperparam_optimization.py

@@ -145,7 +145,9 @@ def objective(self, trial : optuna.trial.Trial, model: QSPRModel) -> float:
         # evaluate the model with the current parameters and return the score
         y, y_ind = model.data.getTargetPropertiesValues()
         score = self.scoreFunc(
-            y, model.evaluate(save=False, parameters=bayesian_params)
+            y, model.evaluate(save=False,
+                              parameters=bayesian_params,
+                              score_func=self.scoreFunc)
         )
         return score
 
@@ -182,7 +184,9 @@ def optimize(self, model: QSPRModel, save_params: bool = True) -> dict:
         for params in ParameterGrid(self.paramGrid):
             logger.info(params)
             y, y_ind = model.data.getTargetPropertiesValues()
-            score = self.scoreFunc(y, model.evaluate(save=False, parameters=params))
+            score = self.scoreFunc(y, model.evaluate(save=False,
+                                                     parameters=params,
+                                                     score_func=self.scoreFunc))
             logger.info("Score: %s" % score)
             if score > self.bestScore:
                 self.bestScore = score

qsprpred/models/metrics.py

@@ -158,6 +158,7 @@ class SklearnMetric(Metric):
         "jaccard_micro",
         "jaccard_macro",
         "jaccard_weighted",
+        'matthews_corrcoef',
     ]
     multiClassMetrics = [
         "neg_log_loss",
@@ -303,6 +304,7 @@ def needsDiscreteToScore(self):
             "recall_macro",
             "recall_weighted",
             "recall_samples",
+            "matthews_corrcoef",
         ]
 
     def supportsTask(self, task: ModelTasks):

qsprpred/models/models.py

@@ -112,6 +112,7 @@ def evaluate(
         self,
         save: bool = True,
         parameters: dict = None,
+        score_func=None,
         **kwargs
     ) -> float | np.ndarray:
         """Make predictions for crossvalidation and independent test set.
@@ -122,6 +123,8 @@ def evaluate(
                 (don't save predictions when used in bayesian optimization)
             parameters (dict):
                 model parameters, if None, the parameters from the model are used
+            score_func (Metric): 
+                metric to use for scoring, if None, the metric from the model is used
             **kwargs:
                 additional keyword arguments for the estimator's predict method
 
@@ -130,6 +133,8 @@ def evaluate(
                 predictions for evaluation
         """
         evalparams = self.parameters if parameters is None else parameters
+        score_func = self.scoreFunc if score_func is None else score_func
+
         # check if data is available
         self.checkForData()
         folds = self.data.createFolds()