Canonicalize SMILES code (#507)

Use RdKit to canonicalize the SMILES code given from the user.
When the canonical SMILES differs from the input, print it.
Apparently, some SMILES failed to generate conformer when they were not canonicalized. 
Closes #505 and #331

aiidalab_widgets_base/structures.py

@@ -772,7 +772,10 @@ def _rdkit_opt(self, smiles, steps):
             return None
         mol = Chem.AddHs(mol)
 
-        AllChem.EmbedMolecule(mol, maxAttempts=20, randomSeed=42)
+        conf_id = AllChem.EmbedMolecule(mol, maxAttempts=20, randomSeed=42)
+        if conf_id < 0:
+            self.output.value = "RDKit ERROR: Could not generate conformer"
+            return None
         if AllChem.UFFHasAllMoleculeParams(mol):
             AllChem.UFFOptimizeMolecule(mol, maxIters=steps)
         else:
@@ -785,8 +788,18 @@ def _rdkit_opt(self, smiles, steps):
 
     def _mol_from_smiles(self, smiles, steps=1000):
         """Convert SMILES to ase structure try rdkit then pybel"""
+
+        # Canonicalize the SMILES code
+        # https://en.wikipedia.org/wiki/Simplified_molecular-input_line-entry_system#Terminology
+        canonical_smiles = self.canonicalize_smiles(smiles)
+        if not canonical_smiles:
+            return None
+
+        if canonical_smiles != smiles:
+            self.output.value = f"Canonical SMILES: {canonical_smiles}"
+
         try:
-            return self._rdkit_opt(smiles, steps)
+            return self._rdkit_opt(canonical_smiles, steps)
         except ValueError as e:
             self.output.value = str(e)
             if self.disable_openbabel:
@@ -802,11 +815,27 @@ def _on_button_pressed(self, change=None):
             return
         spinner = f"Screening possible conformers {self.SPINNER}"  # font-size:20em;
         self.output.value = spinner
+
         self.structure = self._mol_from_smiles(self.smiles.value)
         # Don't overwrite possible error/warning messages
         if self.output.value == spinner:
             self.output.value = ""
 
+    def canonicalize_smiles(self, smiles):
+        from rdkit import Chem
+
+        mol = Chem.MolFromSmiles(smiles, sanitize=True)
+        if mol is None:
+            # Something is seriously wrong with the SMILES code,
+            # just return None and don't attempt anything else.
+            self.output.value = "RDkit ERROR: Invalid SMILES string"
+            return None
+        canonical_smiles = Chem.MolToSmiles(mol, isomericSmiles=True, canonical=True)
+        if not canonical_smiles:
+            self.output.value = "RDkit ERROR: Could not canonicalize SMILES"
+            return None
+        return canonical_smiles
+
     @tl.default("structure")
     def _default_structure(self):
         return None

tests/test_structures.py

@@ -157,6 +157,27 @@ def test_smiles_widget():
     assert widget.structure.get_chemical_formula() == "N2"
 
 
+@pytest.mark.usefixtures("aiida_profile_clean")
+def test_smiles_canonicalization():
+    """Test the SMILES canonicalization via RdKit."""
+    widget = awb.SmilesWidget()
+
+    # Should not change canonical smiles
+    assert widget.canonicalize_smiles("C") == "C"
+
+    # Should canonicalize this
+    canonical = widget.canonicalize_smiles("O=CC=C")
+    assert canonical == "C=CC=O"
+
+    # Should be idempotent
+    assert canonical == widget.canonicalize_smiles(canonical)
+
+    # Regression test for https://github.com/aiidalab/aiidalab-widgets-base/issues/505
+    # Throwing in this non-canonical string should not raise
+    nasty_smiles = "C=CC1=C(C2=CC=C(C3=CC=CC=C3)C=C2)C=C(C=C)C(C4=CC=C(C(C=C5)=CC=C5C(C=C6C=C)=C(C=C)C=C6C7=CC=C(C(C=C8)=CC=C8C(C=C9C=C)=C(C=C)C=C9C%10=CC=CC=C%10)C=C7)C=C4)=C1"
+    widget._rdkit_opt(nasty_smiles, steps=1)
+
+
 @pytest.mark.usefixtures("aiida_profile_clean")
 def test_basic_cell_editor_widget(structure_data_object):
     """Test the `BasicCellEditor`."""