Throw error for "explicit" hydrogens that aren't in molecule graph

docs/releasehistory.md

@@ -16,6 +16,7 @@ Releases follow the `major.minor.micro` scheme recommended by [PEP440](https://w
 ### Bugfixes
 
 - [PR #1971](https://github.com/openforcefield/openff-toolkit/pull/1971): Fixes bug where OpenEyeToolkitWrapper would write coordinate-less PDB atoms if insertion_code or chain_id was an empty string ([Issue #1967](https://github.com/openforcefield/openff-toolkit/issues/1967))
+- [PR #1983](https://github.com/openforcefield/openff-toolkit/pull/1983): Fixes issue with implicit hydrogens slipping though checks to result in radicals later. [Issue #936](https://github.com/openforcefield/openff-toolkit/issues/936) and [Issue #1696](https://github.com/openforcefield/openff-toolkit/issues/1696)
 
 ### Performance improvements
 
@@ -30,6 +31,7 @@ Releases follow the `major.minor.micro` scheme recommended by [PEP440](https://w
 ## 0.16.7
 
 ### Bugfixes
+
 - [PR #1971](https://github.com/openforcefield/openff-toolkit/pull/1971): Fixes bug where OpenEyeToolkitWrapper would write coordinate-less PDB atoms if insertion_code or chain_id was an empty string ([Issue #1967](https://github.com/openforcefield/openff-toolkit/issues/1967))
 
 ### Tests updated

openff/toolkit/_tests/test_toolkits.py

@@ -2250,6 +2250,27 @@ def test_rdkit_from_smiles_hydrogens_are_explicit(self):
         )
         assert offmol.n_atoms == 4
 
+    def test_rdkit_from_smiles_hydrogens_are_explicit_and_in_graph(self):
+        """
+        Test to ensure that RDKitToolkitWrapper.from_smiles has the proper behavior with
+        respect to its hydrogens_are_explicit kwarg and that a CMILES entry has all hydrogens
+        in the graph.
+
+        `Issue #936 <https://github.com/openforcefield/openff-toolkit/issues/936>`,
+        `Issue #1696 <https://github.com/openforcefield/openff-toolkit/issues/1696>`,
+        """
+        toolkit_wrapper = RDKitToolkitWrapper()
+        smiles_impl = "[H][C]([H])([H])[NH+]([H])[C]([H])([H])[H]"
+        with pytest.raises(
+            ValueError,
+            match="the following SMILES as having some nonexplicit hydrogens",
+        ):
+            _ = Molecule.from_smiles(
+                smiles_impl,
+                toolkit_registry=toolkit_wrapper,
+                hydrogens_are_explicit=True,
+            )
+
     @pytest.mark.parametrize("molecule", get_mini_drug_bank(RDKitToolkitWrapper))
     def test_to_inchi(self, molecule):
         """Test, but do not validate, conversion to standard and non-standard InChI"""

openff/toolkit/utils/rdkit_wrapper.py

@@ -1588,6 +1588,24 @@ def from_smiles(
                         f"desired molecule as an RDMol with no implicit hydrogens, and then use "
                         f"Molecule.from_rdkit() to create the desired OFFMol."
                     )
+                # Issue #936 and #1696: GetNumExplicitHs() counts the number of hydrogens that are
+                # explicitly in the string but not present in the graph. This criteria would apply to
+                # hydrogens inside the brackets descibing a heavy atom, e.g. [NH+].
+                # https://www.rdkit.org/docs/Cookbook.html#explicit-valence-and-number-of-hydrogens
+                # This method is also used in GetTotalNumHs, which returns
+                # GetNumImplicitHs() + GetNumExplicitHs() + Optionally the number of
+                # neighboring hydrogens in the molecular graph.
+                # (see the cookbook link above for more details)
+                elif atom.GetNumExplicitHs() != 0:
+                    raise ValueError(
+                        "'hydrogens_are_explicit' was specified as True, but the RDKit wrapper interpreted "
+                        f"the following SMILES as having some nonexplicit hydrogens (e.g., [NH+]): '{smiles}'\n"
+                        "If this SMILES is intended to express all explicit hydrogens in the molecule, then either: \n"
+                        "1) Replace instances of implicit Hs (ex. `[NH+]` with `[N+]([H])` to make "
+                        "the hydrogen explicit, or\n"
+                        "2) construct the desired molecule as an RDMol with no implicit hydrogens, "
+                        "and then use Molecule.from_rdkit() to create the desired OFFMol."
+                    )
 
         molecule = self.from_rdkit(
             rdmol,