Merge branch 'master' of https://github.com/plpla/DeepCCS

core/DeepCCS/model/DeepCCS.py

@@ -69,6 +69,8 @@ def create_model(self):
         """
             Builds a neural net using a set of arguments
             """
+        if len(self.smiles_encoder.converter) == 0 or len(self.adduct_encoder.converter) ==  0:
+            raise ValueError("Encoders must be fit before creating a model.")
         smile_input_layer = Input(shape=(250, len(self.smiles_encoder.converter)), name="smile")
         conv = Conv1D(64, kernel_size=4, activation='relu', kernel_initializer='normal')(smile_input_layer)
 

core/DeepCCS/model/splitter.py

@@ -25,6 +25,10 @@ class SMILESsplitter:
     def split(self, smiles):
         """
         Split a single SMILES using chemical symbols and characters.
+        Two letters chemical symbol that end with a 'c' might not be handled properly.
+        Nitrogen, Sulfur and Oxygen can miss-handled if they are at the begining of an aromatic structure (ex: Coccc)
+        As and Se will be splitted in two caracters if they are found in an aromatic structure.
+        Only Co is seen in the current dataset and it is handled properly. TODO: better splitting.
         :param smiles: The SMILES to split
         :return: A list of chemical symbol/character ordered as
         """
@@ -35,7 +39,6 @@ def split(self, smiles):
                     splitted_smiles.append(k + smiles[j + 1])
                 else:
                     splitted_smiles.append(k)
-
             elif j != 0 and j < len(smiles) - 1:
                 if k.isupper() and smiles[j + 1].islower() and smiles[j + 1] != "c":
                     splitted_smiles.append(k + smiles[j + 1])
@@ -49,4 +52,4 @@ def split(self, smiles):
                     pass
                 else:
                     splitted_smiles.append(k)
-        return splitted_smiles
+        return splitted_smiles

core/DeepCCS/utils.py

@@ -27,7 +27,7 @@
 import h5py as h5
 import sys
 from sklearn.metrics import r2_score, mean_absolute_error, median_absolute_error
-
+from .model.splitter import SMILESsplitter
 
 def filter_data(data_table):
     """
@@ -38,7 +38,8 @@ def filter_data(data_table):
     # Remove smiles that are too long
     logging.debug("{} items before filtering".format(len(data_table)))
     pre_filter = len(data_table)
-    data = data_table[np.array([len(str(i)) for i in data_table["SMILES"]]) <= MAX_SMILES_LENGTH]
+    smiles_splitter = SMILESsplitter()
+    data = data_table[np.array([len(smiles_splitter.split(i)) for i in data_table["SMILES"]]) <= MAX_SMILES_LENGTH]
     # Remove empty smiles
     data = data[np.array([len(str(i)) for i in data["SMILES"]]) > 0]
     data = data.dropna(axis=0, how="any", subset=["SMILES"])