Ensure SD tags are parsed from OEGraphMol

docs/releasehistory.md

@@ -14,6 +14,8 @@ Releases follow the `major.minor.micro` scheme recommended by [PEP440](https://w
 
 ### Bugfixes
 
+- [PR #1778](https://github.com/openforcefield/openff-toolkit/pull/1778): Ensures SD data tags are preserved in `Molecule.from_openeye` if the input is of type `oechem.OEGraphMol`.
+
 ### New features
 
 ### Improved documentation and warnings

openff/toolkit/_tests/test_toolkits.py

@@ -550,6 +550,17 @@ def test_to_from_openeye_core_props_unset(self):
             == expected_output_smiles
         )
 
+    def test_sd_tags_in_graph_mol_preserved(self):
+        from openeye import oechem
+
+        graph = oechem.OEGraphMol()
+
+        oechem.OEParseSmiles(graph, "CCO")
+
+        oechem.OEAddSDData(graph, "cat", "meow")
+
+        assert Molecule.from_openeye(graph).properties["cat"] == "meow"
+
     def test_to_from_openeye_none_partial_charges(self):
         """Test to ensure that to_openeye and from_openeye correctly handle None partial charges"""
         import math

openff/toolkit/topology/molecule.py

@@ -2718,6 +2718,8 @@ def assign_fractional_bond_orders(
     def _invalidate_cached_properties(self) -> None:
         """
         Indicate that the chemical entity has been altered.
+
+        Note that this does not clear the `.properties` dictionary attribute.
         """
         self._conformers = None
         self._partial_charges = None

openff/toolkit/utils/openeye_wrapper.py

@@ -1125,6 +1125,12 @@ def from_openeye(
 
         from openeye import oechem
 
+        # Save the existing SD tags, if any, since they're lost in the cast to
+        # OEMol of the input is OEGraphMol. See issue #1711
+        existing_sd_tags = {
+            pair.GetTag(): pair.GetValue() for pair in oechem.OEGetSDDataIter(oemol)
+        }
+
         oemol = oechem.OEMol(oemol)
 
         # Add explicit hydrogens if they're implicit
@@ -1207,9 +1213,9 @@ def describe_oeatom(oeatom) -> str:
         if oemol.GetTitle() != "":
             molecule.name = oemol.GetTitle()
 
-        # Copy any attached SD tag information
-        for dp in oechem.OEGetSDDataPairs(oemol):
-            molecule._properties[dp.GetTag()] = dp.GetValue()
+        # Attach any SD tag information we saved before casting input to new OEMol
+        for key, value in existing_sd_tags.items():
+            molecule._properties[key] = value
 
         off_to_oe_idx = dict()  # {oemol_idx: molecule_idx}
         atom_mapping = {}