Use the retained azole names instead of spelling the rings out

Hantzsch-Widman retained names cover five-membered rings, but eight of them
had no entry in the monocycle table, so the ring namer fell through to
replacement nomenclature: 2-methyl-1,3,4-thiadiazole came out as
2-methyl-1-thia-3,4-diazacyclopenta-2,4-diene.  That is 4,539 names in a
200k sample, 2.3% of everything generated.

The existing discriminators could not tell the isomers apart.  Both the gap
and the distance multiset are symmetric over all heteroatoms, and 1,2,4- and
1,3,4-oxadiazole share a gap multiset of 1,2,2; so do 1,2,3- and 1,2,5- with
1,1,3.  Measuring from the chalcogen instead separates each pair -- 1,2 vs
2,2, and 1,2 vs 1,1 -- and combined with the gap multiset it identifies all
four isomers of each series uniquely.  The O and S series then only differ
by the symbol counts the specs already carry.

Every one of the 4,539 changed names was checked against OPSIN independently
of the self-audit: 4,516 resolve to their input, and the remaining 23 were
already unparseable or wrong under their old names -- not one name got worse.

Five of the eight were missing from the audit's ring-stem table too, which
cost 862 confirmations until they were added.  A new test asserts the two
tables agree, and it immediately turned up 1,2,3-triazine as a pre-existing
instance of the same omission; adding it recovers 3 more.

zinc22 unchanged at 97.63%, pubchem 96.21% -> 96.22%, refutations 27.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: AdrianM0

src/openclatura/audit/substituent_reconstruction.py

@@ -179,9 +179,13 @@
     "1,2,3-triazole": ("[nH]1nncc1", ["1", "2", "3", "4", "5"]),
     "1,2,4-triazole": ("[nH]1ncnc1", ["1", "2", "3", "4", "5"]),
     "tetrazole": ("[nH]1nnnc1", ["1", "2", "3", "4", "5"]),
+    "1,2,3-oxadiazole": ("o1nncc1", ["1", "2", "3", "4", "5"]),
     "1,2,4-oxadiazole": ("o1ncnc1", ["1", "2", "3", "4", "5"]),
+    "1,2,5-oxadiazole": ("o1nccn1", ["1", "2", "3", "4", "5"]),
     "1,3,4-oxadiazole": ("o1cnnc1", ["1", "2", "3", "4", "5"]),
+    "1,2,3-thiadiazole": ("s1nncc1", ["1", "2", "3", "4", "5"]),
     "1,2,4-thiadiazole": ("s1ncnc1", ["1", "2", "3", "4", "5"]),
+    "1,2,5-thiadiazole": ("s1nccn1", ["1", "2", "3", "4", "5"]),
     "1,3,4-thiadiazole": ("s1cnnc1", ["1", "2", "3", "4", "5"]),
     # benzo-fused five-membered heterocycles (labels 1,2,3,3a,4,5,6,7,7a)
     "indole": ("[nH]1ccc2ccccc21", ["1", "2", "3", "3a", "4", "5", "6", "7", "7a"]),
@@ -226,6 +230,7 @@
     "1,2-oxazolidine": ("O1NCCC1", ["1", "2", "3", "4", "5"]),
     # six-membered triazines
     "1,3,5-triazine": ("n1cncnc1", ["1", "2", "3", "4", "5", "6"]),
+    "1,2,3-triazine": ("n1nnccc1", ["1", "2", "3", "4", "5", "6"]),
     "1,2,4-triazine": ("n1ncncc1", ["1", "2", "3", "4", "5", "6"]),
     # benzo-fused carbocycle / 2,3-dihydro azole (labels 1,2,3,3a,4,5,6,7,7a)
     "indoline": ("N1CCc2ccccc21", ["1", "2", "3", "3a", "4", "5", "6", "7", "7a"]),

src/openclatura/data/namer_rules.json

@@ -2016,6 +2016,158 @@
           2
         ],
         "no_cumulated_double_bonds": true
+      },
+      {
+        "name": "1,2,3-oxadiazole",
+        "size": 5,
+        "double_bonds": 2,
+        "symbols": {
+          "N": 2,
+          "O": 1
+        },
+        "hetero_gap_multiset": [
+          1,
+          1,
+          3
+        ],
+        "chalcogen_nitrogen_distance_multiset": [
+          1,
+          2
+        ],
+        "no_cumulated_double_bonds": true
+      },
+      {
+        "name": "1,2,4-oxadiazole",
+        "size": 5,
+        "double_bonds": 2,
+        "symbols": {
+          "N": 2,
+          "O": 1
+        },
+        "hetero_gap_multiset": [
+          1,
+          2,
+          2
+        ],
+        "chalcogen_nitrogen_distance_multiset": [
+          1,
+          2
+        ],
+        "no_cumulated_double_bonds": true
+      },
+      {
+        "name": "1,2,5-oxadiazole",
+        "size": 5,
+        "double_bonds": 2,
+        "symbols": {
+          "N": 2,
+          "O": 1
+        },
+        "hetero_gap_multiset": [
+          1,
+          1,
+          3
+        ],
+        "chalcogen_nitrogen_distance_multiset": [
+          1,
+          1
+        ],
+        "no_cumulated_double_bonds": true
+      },
+      {
+        "name": "1,3,4-oxadiazole",
+        "size": 5,
+        "double_bonds": 2,
+        "symbols": {
+          "N": 2,
+          "O": 1
+        },
+        "hetero_gap_multiset": [
+          1,
+          2,
+          2
+        ],
+        "chalcogen_nitrogen_distance_multiset": [
+          2,
+          2
+        ],
+        "no_cumulated_double_bonds": true
+      },
+      {
+        "name": "1,2,3-thiadiazole",
+        "size": 5,
+        "double_bonds": 2,
+        "symbols": {
+          "N": 2,
+          "S": 1
+        },
+        "hetero_gap_multiset": [
+          1,
+          1,
+          3
+        ],
+        "chalcogen_nitrogen_distance_multiset": [
+          1,
+          2
+        ],
+        "no_cumulated_double_bonds": true
+      },
+      {
+        "name": "1,2,4-thiadiazole",
+        "size": 5,
+        "double_bonds": 2,
+        "symbols": {
+          "N": 2,
+          "S": 1
+        },
+        "hetero_gap_multiset": [
+          1,
+          2,
+          2
+        ],
+        "chalcogen_nitrogen_distance_multiset": [
+          1,
+          2
+        ],
+        "no_cumulated_double_bonds": true
+      },
+      {
+        "name": "1,2,5-thiadiazole",
+        "size": 5,
+        "double_bonds": 2,
+        "symbols": {
+          "N": 2,
+          "S": 1
+        },
+        "hetero_gap_multiset": [
+          1,
+          1,
+          3
+        ],
+        "chalcogen_nitrogen_distance_multiset": [
+          1,
+          1
+        ],
+        "no_cumulated_double_bonds": true
+      },
+      {
+        "name": "1,3,4-thiadiazole",
+        "size": 5,
+        "double_bonds": 2,
+        "symbols": {
+          "N": 2,
+          "S": 1
+        },
+        "hetero_gap_multiset": [
+          1,
+          2,
+          2
+        ],
+        "chalcogen_nitrogen_distance_multiset": [
+          2,
+          2
+        ],
+        "no_cumulated_double_bonds": true
       }
     ]
   },

src/openclatura/rules/retained.py

@@ -941,6 +941,11 @@ def _match_data_monocycle_retained(
         expected_gaps = spec.get("hetero_gap_multiset")
         if expected_gaps is not None and _hetero_gap_multiset(mol, path) != sorted(expected_gaps):
             continue
+        expected_chalcogen = spec.get("chalcogen_nitrogen_distance_multiset")
+        if expected_chalcogen is not None and _chalcogen_nitrogen_distance_multiset(mol, path) != sorted(
+            expected_chalcogen
+        ):
+            continue
         return spec["name"]
     return None
 
@@ -1031,6 +1036,21 @@ def _hetero_distance_multiset(mol: Molecule, path: list[int]) -> list[int]:
     return sorted(distances)
 
 
+def _chalcogen_nitrogen_distance_multiset(mol: Molecule, path: list[int]) -> list[int]:
+    """Ring distances from each chalcogen to each nitrogen.
+
+    The gap multiset is symmetric over all heteroatoms, so it cannot tell
+    1,2,4-oxadiazole from 1,3,4-oxadiazole -- both are gaps 1,2,2.  Measuring
+    from the chalcogen separates them: 1 and 2 against 2 and 2.
+    """
+
+    size = len(path)
+    chalcogens = [idx for idx, atom_idx in enumerate(path) if mol.atoms[atom_idx].symbol in {"O", "S"}]
+    nitrogens = [idx for idx, atom_idx in enumerate(path) if mol.atoms[atom_idx].symbol == "N"]
+    distances = [min(abs(left - right), size - abs(left - right)) for left in chalcogens for right in nitrogens]
+    return sorted(distances)
+
+
 def _hetero_gap_multiset(mol: Molecule, path: list[int]) -> list[int]:
     hetero_indices = sorted(idx for idx, atom_idx in enumerate(path) if mol.atoms[atom_idx].symbol in {"N", "O", "S"})
     if len(hetero_indices) < 3:

src/openclatura/tests/test_rules_tables.py

@@ -140,3 +140,42 @@ def test_uncitable_added_hydrogen_keeps_the_von_baeyer_parent():
     name = name_smiles("N#CC[C@H](O)CNC(=O)c1c[nH]c2ccccc2c1=O")
     assert "quinoline" not in name
     assert "bicyclo[4.4.0]" in name
+
+
+def test_oxadiazoles_and_thiadiazoles_use_their_retained_names():
+    """Replacement nomenclature spelt these out as 1-thia-3,4-diazacyclopenta-
+    2,4-diene; Hantzsch-Widman retained names cover 5-rings, so they should not."""
+
+    from openclatura import name_smiles
+
+    cases = {
+        "c1cnno1": "1,2,3-oxadiazole",
+        "c1ncno1": "1,2,4-oxadiazole",
+        "c1nonc1": "1,2,5-oxadiazole",
+        "c1nnco1": "1,3,4-oxadiazole",
+        "c1cnns1": "1,2,3-thiadiazole",
+        "c1ncns1": "1,2,4-thiadiazole",
+        "c1nsnc1": "1,2,5-thiadiazole",
+        "c1nncs1": "1,3,4-thiadiazole",
+    }
+    for smiles, expected in cases.items():
+        assert name_smiles(smiles) == expected, smiles
+
+    # The isomers must stay distinguishable once substituted: a symmetric gap
+    # multiset cannot separate 1,2,4- from 1,3,4-, which is what the chalcogen
+    # distance criterion is for.
+    assert name_smiles("Cc1nnc(C)s1") == "2,5-dimethyl-1,3,4-thiadiazole"
+    assert name_smiles("Cc1ncns1") == "5-methyl-1,2,4-thiadiazole"
+    assert name_smiles("Cc1nnc(-c2ccccc2)o1") == "2-methyl-5-phenyl-1,3,4-oxadiazole"
+
+
+def test_audit_models_every_retained_monocycle_the_namer_emits():
+    """A retained ring the audit has no template for abstains, so enabling one
+    without the matching template silently stops it being verified."""
+
+    from openclatura.audit.reconstruction import _ALL_PARENT_TEMPLATES
+    from openclatura.nomenclature import RULES
+
+    emitted = {spec["name"] for spec in RULES.retained.monocycle_specs}
+    missing = sorted(name for name in emitted if name not in _ALL_PARENT_TEMPLATES)
+    assert missing == []