[Feature] Management Command to Compute and Store Reaction <-> Rule Associations (#453)

Co-authored-by: Tim Lorsbach <tim@lorsba.ch>
Reviewed-on: enviPath/enviPy#453

.gitea/workflows/build-image.yaml

@@ -0,0 +1,62 @@
+name: Build Docker Image
+
+# Trigger when a PR to main/develop is completed.
+on:
+  pull_request:
+    branches:
+      - main
+      - develop
+    types:
+      - closed
+
+jobs:
+  build-and-push:
+    if: ${{ github.event.pull_request.merged == true }}
+    runs-on: ubuntu-latest
+    steps:
+      # Fetch the repository content for the Docker build context.
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      # Enable Buildx for BuildKit features (incl. SSH mount support).
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      # Authenticate against the container registry before pushing images.
+      - name: Log in to container registry
+        uses: docker/login-action@v3
+        with:
+          registry: git.envipath.com
+          username: ${{ secrets.CI_REGISTRY_USER }}
+          password: ${{ secrets.CI_REGISTRY_PASSWORD }}
+
+      # Generate image tags/labels:
+      # - PRs targeting main get "latest" and "main-sha"
+      # - PRs targeting develop get "dev" and "dev-sha"
+      - name: Extract metadata
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          images: git.envipath.com/envipath/envipy
+          tags: |
+            type=raw,value=latest,enable=${{ github.event.pull_request.base.ref == 'main' }}
+            type=sha,prefix=main-,enable=${{ github.event.pull_request.base.ref == 'main' }}
+            type=raw,value=dev,enable=${{ github.event.pull_request.base.ref == 'develop' }}
+            type=sha,prefix=dev-,enable=${{ github.event.pull_request.base.ref == 'develop' }}
+
+      # Load SSH key so Docker can pull private git+ssh dependencies during build.
+      - name: Setup SSH for private git dependencies
+        uses: webfactory/ssh-agent@v0.9.0
+        with:
+          ssh-private-key: ${{ secrets.ENVIPY_CI_PRIVATE_KEY }}
+
+      # Build and push the production image; forward SSH agent without registry cache reuse.
+      - name: Build and push Docker image
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          file: Dockerfile
+          push: true
+          ssh: default
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}

Dockerfile

@@ -58,6 +58,10 @@ COPY tests tests
 COPY utilities utilities
 COPY manage.py .
 
+# Used to run migrations etc
+COPY entrypoint.sh entrypoint.sh
+RUN chmod +x entrypoint.sh
+
 # Install frontend deps
 COPY package.json pnpm-lock.yaml pnpm-workspace.yaml ./
 
@@ -100,4 +104,5 @@ USER django
 
 EXPOSE 8000
 
-CMD ["gunicorn", "envipath.wsgi:application", "--bind", "0.0.0.0:8000", "--workers", "3"]
+ENTRYPOINT ["/app/entrypoint.sh"]
+CMD ["gunicorn", "envipath.wsgi:application", "--bind", "0.0.0.0:8000", "--workers", "8"]

entrypoint.sh

@@ -0,0 +1,9 @@
+#!/bin/bash
+set -e
+
+if [ "${SKIP_DJANGO_SETUP:-false}" != "true" ]; then
+  python manage.py migrate --no-input
+  python manage.py collectstatic --no-input
+fi
+
+exec "$@"

epdb/management/commands/localize_urls.py

@@ -44,15 +44,19 @@ class Command(BaseCommand):
             "EPModel",
             "ApplicabilityDomain",
             "EnzymeLink",
+            "AdditionalInformation",
         ]
         for model in MODELS:
             obj_cls = apps.get_model("epdb", model)
-            obj_cls.objects.update(
-                url=Replace(F("url"), Value(options["old"]), Value(options["new"]))
+
+            update_fields = {"url": Replace(F("url"), Value(options["old"]), Value(options["new"]))}
+            if hasattr(obj_cls, "description"):
+                update_fields["description"] = Replace(
+                    F("description"), Value(options["old"]), Value(options["new"])
                 )
+
             if issubclass(obj_cls, EnviPathModel):
-                obj_cls.objects.update(
-                    kv=Cast(
+                update_fields["kv"] = Cast(
                     Replace(
                         Cast(F("kv"), output_field=TextField()),
                         Value(options["old"]),
@@ -60,4 +64,5 @@ class Command(BaseCommand):
                     ),
                     output_field=JSONField(),
                 )
-                )
+
+            obj_cls.objects.update(**update_fields)

epdb/management/commands/reaction_rule_mapping.py

@@ -0,0 +1,97 @@
+import logging
+
+from django.core.management.base import BaseCommand
+from django.db import transaction
+from uuid import uuid4
+from epdb.models import Package, ReactionExplanation
+from utilities.chem import FormatConverter
+from django.utils import timezone
+
+
+logger = logging.getLogger(__name__)
+
+
+class Command(BaseCommand):
+    def add_arguments(self, parser):
+        parser.add_argument(
+            "--rule-package",
+            action="append",
+            default=["32de3cf4-e3e6-4168-956e-32fa5ddb0ce1"],
+            type=str,
+            help="UUID to process. Can be specified multiple times.",
+        )
+
+        parser.add_argument(
+            "--reaction-package",
+            action="append",
+            default=[
+                "32de3cf4-e3e6-4168-956e-32fa5ddb0ce1",  # BBD
+                "f05e38d8-e9b4-4c3e-b0d8-9ab29966eccf",  # Sediment
+                "521c547a-fd2a-491c-ad5b-7eaa1577fb65",  # Sludge
+                "5882df9c-dae1-4d80-a40e-db4724271456",  # Soil
+                "87a49584-d937-482c-9c33-25928dcb02a8",  # PFAS
+            ],
+            type=str,
+            help="UUID to process. Can be specified multiple times.",
+        )
+
+        parser.add_argument(
+            "--dry-run",
+            default=False,
+            action="store_true",
+            help="Perform dry run",
+        )
+
+    @transaction.atomic
+    def handle(self, *args, **options):
+        RUN_UUID = uuid4()
+        RUN_START = timezone.now()
+
+        rule_packages = Package.objects.filter(uuid__in=options["rule_package"])
+        reaction_packages = Package.objects.filter(uuid__in=options["reaction_package"])
+
+        rules = []
+        for rule_package in rule_packages:
+            rules.extend(rule_package.get_applicable_rules())
+
+        reactions = []
+        for reaction_package in reaction_packages:
+            reactions.extend(reaction_package.reactions)
+
+        logger.debug(f"Collected {len(rules)} rules and {len(reactions)} reactions.")
+
+        for i, reaction in enumerate(reactions):
+            logger.debug(f"Reaction {i} / {len(reactions)}")
+            for j, rule in enumerate(rules):
+                reactants, products = reaction.smirks().split(">>")
+
+                if len(reactants.split(".")) > 1:
+                    logger.debug(f"Skipping reaction {reaction.uuid} as it has multiple reactants.")
+                    break
+
+                products = products.split(".")
+
+                # Run reaction with rule
+                rule_products = rule.apply(reactants)
+
+                # Check if products match (in both directions if extras are not allowed)
+                for product_set in rule_products:
+                    covered, exact = FormatConverter.smiles_covered_by(
+                        products,
+                        product_set.product_set,
+                        standardize=True,
+                        canonicalize_tautomers=True,
+                        return_exact_match=True,
+                    )
+
+                    if covered and not options["dry-run"]:
+                        logger.debug(f"Reaction {reaction.uuid} explained by rule {rule.uuid}")
+                        re = ReactionExplanation()
+                        re.run_uuid = RUN_UUID
+                        re.run_start = RUN_START
+                        re.reaction = reaction
+                        re.rule = rule
+                        re.exact = exact
+                        re.save()
+                        # Its explained, if there are more sets skip them
+                        break

epdb/migrations/0028_auto_20260812_0902.py

@@ -0,0 +1,113 @@
+# Generated by Django 6.0.3 on 2026-08-12 09:02
+
+from django.conf import settings as s
+from django.db import migrations
+from envipy_additional_information import Likelihood, RuleLikelihood
+
+NEW_RULE = {
+    "parent": "bt0005",
+    "name": "bt0005-3667",
+    "description": "vic-unsubstituted Aromatic > vic-Dihydroxyaromatic",
+    "smirks": "[#8:7]([H])-[#6:1]([H])-1-[#6:2]=[#6:3]-[#6:4]=[#6:5]-[#6:6]([H])-1-[#8:8]([H])>>[#8:7]([H])-[#6:1]=1-[#6:2]=[#6:3]-[#6:4]=[#6:5]-[#6:6]=1-[#8:8]([H])",
+    "scenario_name": "bt0005-3667 aerobic likelihood",
+    "scenario_aerobic_likelihood": RuleLikelihood(likelihood=Likelihood.LIKELY),
+}
+
+RULE_FIXES = {
+    "bt0005-4282": "[c:1]([H])1:[c:2]([H]):[#6,#7;a:3]:[c:4]:[c:5]:[c:6]1>>[c:1]([#8])1:[c:2]([#8]):[#6,#7;a:3]:[c:4]:[c:5]:[c:6]1",
+    "bt0014-4215": "[c:1]([H])1[c:8][#6,#7;a:7][c:6][c:5][c:4]1[#8;!$([OH]c:[#6,#7;a:7]([OH])):9]([H])>>[#8:9]([H])[c:4]1:[c:5]:[c:6]:[#6,#7;a:7]:[c:8]:[c:1]1[#8]([H])",
+    # "bt0063-3938": "[#1,#6:6][#7;X3;!$(NC1CC1)!$([N][C]=O)!$([!#8]CNC=O):1]([#1,#6:7])[#6;A;X4:2][H:3]>>[#1,#6:6][#7;X3:1]([H:3])(=[#1,#6:7]).[#6;A:2]=O",
+    # CN1C=NC2=C1C(=O)N(C)C(=O)N2 not working anymore with bt0063-3938 if change above is applied
+    "bt0063-3938": "[#1,#6:6][#7;X3;!$(NC1CC1)!$([N][C]=O)!$([!#8]CNC=O):1]([#1,#6:7])[#6;A;X4:2][H:3]>>[#1,#6:6][#7;X3:1]([#1,#6:7])[H:3].[#6;A:2]=O",
+    "bt0068-3564": "[#7:4]!@-[#6:2](!@-[#7:1])=[O:5]>>[#7:4]-[#6:2](-[O+0H1])=[O:5].[#7H1:1]",
+    "bt0180-2844": "[H][C:2]([#6:5]([H])([H])([H]))([#1,#6:4])!@-[#6:1]([H])([H])-[#6:3](-[#8-:8])=[O:6]>>[#6:5]([H])([H])([H])\\[#6:2](-[#1,#6:4])=[#6H:1]\\[#6:3](-[#8-:8])=[O:6]",
+    "bt0181-1278": "[#8-:1]-[#6:2](=[O:11])-[#6:7]=[#6:8]-[#6:3](-[H])=[#6:5](-Cl)-[#6:6](-[#8-:10])=[O:9]>>[O+0H1:10]-[#6:6](=[O:9])-[#6:5]=[#6:3]-1-[O+0:1]-[#6:2](=[O:11])-[#6:7]=[#6:8]-1",
+    "bt0298-3335": "[#6:1][N+:2]#[C:3]>>[#6:1]-[#7H2:2]-[#6:3]=O",
+    "bt0322-3393": "[H:10]\\[#6:6](=[#6:9](/[#6:1]([H])([H])([H]))-[#6:11]-[#6:12]-[#6:13]=[#6:14])-[#6:5](-[#16:7])=[O:8]>>[H:10]\\[#6:6](-[#6:5](-[#16:7])=[O:8])=[#6:9](\\[#6:11]-[#6:12]-[#6:13]=[#6:14])-[#6:1]-[#6](-[#8-])=O",
+    "bt0343-2675": "[#8-]-[#6](=O)-[c:1]1[c:6][cH:7][c:8](-[#7H2,#8H1:9])[cH:10][c:11]1>>[#8H][c:1]1[c:6][c:7][c:8]([*:9])[c:10][c:11]1",
+    "bt0350-3319": "[#6:6][#7:3][#6;!R:2]=[#7;!R:1][#6:5]>>[#6:5][#7:1][#6:2]=O.[#6:6][#7:3]",  # Trig before 5 -> all of them shouldn't
+    "bt0374-4081": "[cH:4]1[c:16][c:15][c:14][c:13][c:3]1[#7,#8:2][c:1]1[c:8][c:9][c:10][c:11][c:12]1>>[#7,#8:2]-[c:1]1[c:12][c:11][c:10][c:9][c:8]1[c:13]1[c:14][c:15][c:16][c:4](-[#8])[c:3]1-[#8]",
+    "bt0378-3188": "[#8-:7][c:1]1[c:6]([#7+]([#8-])=O)[c:5][c:4]([#7+:9]([#8-])=O)[c:3][c:2]1([#7+:8]([#8-])=O)>>[#8+0:7]=[#6:1]1-[#6:6]-[#6:5]-[#6:4]([#7+:9]([#8-])=O)-[#6:3]-[#6:2]1([#7+:8]([#8-])=O)",
+    "bt0379-3190": "[#9,#17,#35,#53]-[#6:1](-[H])-1-[#6:5]-,=[#6:6]-[#6:7]-,=[#6:8]-[#6:2](-[H])-1-[#9,#17,#35,#53]>>[#6:6]~1-[#6:7]~[#6:8]-[#6:2]=[#6:1]-[#6:5]~1",
+    "bt0393-3367": "[#6:5]-[#6:1](-[#7:2](-[H])(-[H]))=[S+:3]-[#8-:6]>>[#6:5]-[#6:1](=[#7H1:2])-[S+0:3](=[O])-[#8+0H1:6]",
+}
+
+
+def forward_func(apps, schema_editor):
+    ContentType = apps.get_model("contenttypes", "ContentType")
+
+    pkg_class = s.EPDB_PACKAGE_MODEL
+
+    if len(pkg_class.split(".")) != 2:
+        raise ValueError(
+            f"EPDB_PACKAGE_MODEL must be of the form 'app_label.model_name', got {pkg_class}"
+        )
+
+    app_label, model_name = pkg_class.split(".")
+    Package = apps.get_model(app_label, model_name)
+    SimpleAmbitRule = apps.get_model("epdb", "SimpleAmbitRule")
+    ParallelRule = apps.get_model("epdb", "ParallelRule")
+    Scenario = apps.get_model("epdb", "Scenario")
+    AdditionalInformation = apps.get_model("epdb", "AdditionalInformation")
+    simple_ambit_rule_ct = ContentType.objects.get_for_model(SimpleAmbitRule)
+
+    if Package.objects.filter(name="EAWAG-BBD").exists():
+        p = Package.objects.get(name="EAWAG-BBD")
+
+        if not SimpleAmbitRule.objects.filter(package=p, name=NEW_RULE["name"]).exists():
+            # Create Missing Rule
+            new_sr = SimpleAmbitRule()
+            new_sr.polymorphic_ctype = simple_ambit_rule_ct
+            new_sr.package = p
+            new_sr.name = NEW_RULE["name"]
+            new_sr.description = NEW_RULE["description"]
+            new_sr.smirks = NEW_RULE["smirks"]
+            new_sr.save()
+
+            new_sr.url = "{}/simple-ambit-rule/{}".format(new_sr.package.url, new_sr.uuid)
+            new_sr.save()
+
+            # Add likelihood
+            new_scen = Scenario()
+            new_scen.package = p
+            new_scen.name = NEW_RULE["scenario_name"]
+            new_scen.save()
+
+            new_scen.url = "{}/scenario/{}".format(new_scen.package.url, new_scen.uuid)
+            new_scen.save()
+
+            ai = NEW_RULE["scenario_aerobic_likelihood"]
+            new_add_inf = AdditionalInformation()
+            new_add_inf.package = p
+            new_add_inf.type = ai.__class__.__name__
+            new_add_inf.data = ai.model_dump(mode="json")
+            new_add_inf.scenario = new_scen
+            new_add_inf.save()
+
+            new_add_inf.url = "{}/additional-information/{}".format(
+                new_add_inf.scenario.url, new_add_inf.uuid
+            )
+            new_add_inf.save()
+
+            # Link Scenario
+            new_sr.scenarios.add(new_scen)
+
+            # Link to bt0005
+            pr = ParallelRule.objects.get(package=p, name="bt0005")
+            pr.simple_rules.add(new_sr)
+
+        # Update others
+        for rule_name, smirks in RULE_FIXES.items():
+            sr = SimpleAmbitRule.objects.get(package=p, name=rule_name)
+            sr.smirks = smirks
+            sr.save()
+
+
+class Migration(migrations.Migration):
+    dependencies = [
+        ("epdb", "0027_alter_compound_aliases_and_more"),
+    ]
+
+    operations = [
+        migrations.RunPython(forward_func, reverse_code=migrations.RunPython.noop),
+    ]

epdb/migrations/0029_reactionexplanation_reaction_explained_by.py

@@ -0,0 +1,63 @@
+# Generated by Django 6.0.3 on 2026-08-13 09:58
+
+import django.db.models.deletion
+import django.utils.timezone
+import model_utils.fields
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+    dependencies = [
+        ("epdb", "0028_auto_20260812_0902"),
+    ]
+
+    operations = [
+        migrations.CreateModel(
+            name="ReactionExplanation",
+            fields=[
+                (
+                    "id",
+                    models.BigAutoField(
+                        auto_created=True, primary_key=True, serialize=False, verbose_name="ID"
+                    ),
+                ),
+                (
+                    "created",
+                    model_utils.fields.AutoCreatedField(
+                        default=django.utils.timezone.now, editable=False, verbose_name="created"
+                    ),
+                ),
+                (
+                    "modified",
+                    model_utils.fields.AutoLastModifiedField(
+                        default=django.utils.timezone.now, editable=False, verbose_name="modified"
+                    ),
+                ),
+                ("run_uuid", models.UUIDField()),
+                ("run_start", models.DateTimeField()),
+                ("exact", models.BooleanField(default=False)),
+                (
+                    "reaction",
+                    models.ForeignKey(
+                        on_delete=django.db.models.deletion.CASCADE, to="epdb.reaction"
+                    ),
+                ),
+                (
+                    "rule",
+                    models.ForeignKey(on_delete=django.db.models.deletion.CASCADE, to="epdb.rule"),
+                ),
+            ],
+            options={
+                "abstract": False,
+            },
+        ),
+        migrations.AddField(
+            model_name="reaction",
+            name="explained_by",
+            field=models.ManyToManyField(
+                related_name="explained_reactions",
+                through="epdb.ReactionExplanation",
+                to="epdb.rule",
+            ),
+        ),
+    ]

epdb/migrations/0030_auto_20260814_0741.py

@@ -0,0 +1,37 @@
+# Generated by Django 6.0.3 on 2026-08-14 07:41
+
+from django.db import migrations
+
+
+def forward_func(apps, schema_editor):
+    ContentType = apps.get_model("contenttypes", "ContentType")
+    AdditionalInformation = apps.get_model("epdb", "AdditionalInformation")
+
+    models = {}
+
+    for c in ContentType.objects.all():
+        try:
+            models[(c.app_label, c.model)] = apps.get_model(c.app_label, c.model)
+        except Exception:
+            pass
+
+    for ai in AdditionalInformation.objects.all():
+        if ai.url is None:
+            if ai.content_type is None:
+                ai.url = "{}/additional-information/{}".format(ai.scenario.url, ai.uuid)
+            else:
+                model = models[(ai.content_type.app_label, ai.content_type.model)]
+                obj = model.objects.get(pk=ai.object_id)
+                ai.url = "{}/additional-information/{}".format(obj.url, ai.uuid)
+
+            ai.save()
+
+
+class Migration(migrations.Migration):
+    dependencies = [
+        ("epdb", "0029_reactionexplanation_reaction_explained_by"),
+    ]
+
+    operations = [
+        migrations.RunPython(forward_func, reverse_code=migrations.RunPython.noop),
+    ]

epdb/models.py

@@ -854,9 +854,15 @@ class Compound(
     @property
     def related_reactions(self):
         return (
+            (
                 Reaction.objects.filter(package=self.package, educts__in=[self.default_structure])
-            | Reaction.objects.filter(package=self.package, products__in=[self.default_structure])
-        ).order_by("name")
+                | Reaction.objects.filter(
+                    package=self.package, products__in=[self.default_structure]
+                )
+            )
+            .distinct()
+            .order_by("name")
+        )
 
     @property
     def related_nodes(self):
@@ -1724,6 +1730,14 @@ class SequentialRuleOrdering(models.Model):
     order_index = models.IntegerField(null=False, blank=False)
 
 
+class ReactionExplanation(TimeStampedModel):
+    run_uuid = models.UUIDField(null=False, blank=False)
+    run_start = models.DateTimeField(null=False, blank=False)
+    reaction = models.ForeignKey("epdb.Reaction", on_delete=models.CASCADE)
+    rule = models.ForeignKey("epdb.Rule", on_delete=models.CASCADE)
+    exact = models.BooleanField(default=False)
+
+
 class Reaction(
     EnviPathModel, AliasMixin, ScenarioMixin, ReactionIdentifierMixin, AdditionalInformationMixin
 ):
@@ -1749,6 +1763,12 @@ class Reaction(
 
     external_identifiers = GenericRelation("ExternalIdentifier")
 
+    explained_by = models.ManyToManyField(
+        "epdb.Rule",
+        through="ReactionExplanation",
+        related_name="explained_reactions",
+    )
+
     def _url(self):
         return "{}/reaction/{}".format(self.package.url, self.uuid)
 
@@ -2574,7 +2594,7 @@ class Node(EnviPathModel, AliasMixin, ScenarioMixin, AdditionalInformationMixin)
                 collected[str(ai.scenario.uuid)]["proposed"] = True
 
             if ai.type == "Confidence":
-                collected[str(ai.scenario.uuid)]["Confidence"] = ai.get().level
+                collected[str(ai.scenario.uuid)]["Confidence"] = ai.get().level.value
 
             if ai.type == "TransformationProductImportance":
                 collected[str(ai.scenario.uuid)]["Transformation product importance"] = (

epdb/views.py

@@ -1080,13 +1080,17 @@ def package_model(request, package_uuid, model_uuid):
                 for pr in pred_res:
                     if len(pr) > 0:
                         products = []
+
                         for prod_set in pr.product_sets:
                             logger.debug(f"Checking {prod_set}")
                             products.append(tuple([x for x in prod_set]))
 
+                        products = list(set(products))
+
+                        for prod in products:
                             res["pred"].append(
                                 {
-                                "products": list(set(products)),
+                                    "products": list(prod),
                                     "probability": pr.probability,
                                     "btrule": {k: getattr(pr.rule, k) for k in ["url", "name"]}
                                     if pr.rule is not None

templates/objects/model/classification_model.html

@@ -109,12 +109,12 @@
         res += "<th scope='row'>" + cnt + "</th>";
         res +=
           "<th scope='row'>" +
-          data[transformation]["products"][0].join(", ") +
+          data[transformation]["products"].join(", ") +
           "</th>";
         res +=
           "<th scope='row'>" +
           "<img width='400' src='{% url 'depict' %}?smiles=" +
-          encodeURIComponent(data[transformation]["products"][0].join(".")) +
+          encodeURIComponent(data[transformation]["products"].join(".")) +
           "'></th>";
         res +=
           "<th scope='row'>" +

tests/views/test_model_views.py

@@ -47,7 +47,7 @@ class ModelViewTest(TestCase):
 
         expected = [
             {
-                "products": [["CCN(CC)C(=O)C1=CC(C=O)=CC=C1"]],
+                "products": ["CCN(CC)C(=O)C1=CC(C=O)=CC=C1"],
                 "probability": 0.75,
                 "btrule": {
                     "url": "http://localhost:8000/package/1869d3f0-60bb-41fd-b6f8-afa75ffb09d3/simple-ambit-rule/2f2e0c39-e109-4836-959f-2bda2524f022",
@@ -55,7 +55,7 @@ class ModelViewTest(TestCase):
                 },
             },
             {
-                "products": [["O=C(O)C1=CC(CO)=CC=C1", "CCNCC"]],
+                "products": ["O=C(O)C1=CC(CO)=CC=C1", "CCNCC"],
                 "probability": 0.25,
                 "btrule": {
                     "url": "http://localhost:8000/package/1869d3f0-60bb-41fd-b6f8-afa75ffb09d3/simple-ambit-rule/0e6e9290-b658-4450-b291-3ec19fa19206",
@@ -63,7 +63,7 @@ class ModelViewTest(TestCase):
                 },
             },
             {
-                "products": [["CCNC(=O)C1=CC(CO)=CC=C1", "CC=O"]],
+                "products": ["CCNC(=O)C1=CC(CO)=CC=C1", "CC=O"],
                 "probability": 0.0,
                 "btrule": {
                     "url": "http://localhost:8000/package/1869d3f0-60bb-41fd-b6f8-afa75ffb09d3/simple-ambit-rule/27a3a353-0b66-4228-bd16-e407949e90df",

utilities/chem.py

@@ -68,6 +68,8 @@ class PredictionResult(object):
 
 
 class FormatConverter(object):
+    tautomer_enumerator = rdMolStandardize.TautomerEnumerator()
+
     @staticmethod
     def mass(smiles):
         return Descriptors.MolWt(FormatConverter.from_smiles(smiles))
@@ -240,8 +242,9 @@ class FormatConverter(object):
             Chem.RemoveStereochemistry(res_mol)
 
         if canonicalize_tautomers:
-            te = rdMolStandardize.TautomerEnumerator()  # idem
-            res_mol = te.Canonicalize(res_mol)
+            tautomers = FormatConverter.tautomer_enumerator.Enumerate(res_mol)
+            if len(tautomers) >= 1:
+                res_mol = FormatConverter.tautomer_enumerator.PickCanonical(tautomers)
 
         return Chem.MolToSmiles(res_mol, kekuleSmiles=True)
 
@@ -389,7 +392,7 @@ class FormatConverter(object):
                                 prods.append(p)
 
                         except ValueError as e:
-                            logger.error(f"Sanitizing and converting failed:\n{e}")
+                            logger.debug(f"Sanitizing and converting failed:\n{e}")
                             continue
 
                     if len(prods):
@@ -397,7 +400,8 @@ class FormatConverter(object):
                         pss.add(ps)
 
         except Exception as e:
-            logger.error(f"Applying {smirks} on {smiles} failed:\n{e}")
+            logger.debug(f"Applying {smirks} on {smiles} failed:\n{e}")
+            pass
 
         return list(pss)
 
@@ -444,6 +448,7 @@ class FormatConverter(object):
         r_smiles: List[str],
         standardize: bool = True,
         canonicalize_tautomers: bool = True,
+        return_exact_match: bool = False,
     ) -> bool:
         """
         Check if all SMILES in the left list are covered by (contained in) the right list.
@@ -482,9 +487,11 @@ class FormatConverter(object):
         if standardize:
             for smi in l_smiles:
                 try:
-                    smi = FormatConverter.standardize(
+                    smi = FormatConverter.canonicalize(
+                        FormatConverter.standardize(
                             smi, remove_stereo=True, canonicalize_tautomers=canonicalize_tautomers
                         )
+                    )
                 except Exception:
                     # :shrug:
                     # logger.debug(f'Standardizing SMILES failed for {smi}')
@@ -497,8 +504,12 @@ class FormatConverter(object):
         if standardize:
             for smi in r_smiles:
                 try:
-                    smi = FormatConverter.standardize(
-                        smi, remove_stereo=True, canonicalize_tautomers=canonicalize_tautomers
+                    smi = FormatConverter.canonicalize(
+                        FormatConverter.standardize(
+                            smi,
+                            remove_stereo=True,
+                            canonicalize_tautomers=canonicalize_tautomers,
+                        )
                     )
                 except Exception:
                     # :shrug:
@@ -507,8 +518,11 @@ class FormatConverter(object):
                 standardized_r_smiles.append(smi)
         else:
             standardized_r_smiles = r_smiles
-
+        if not return_exact_match:
             return len(set(standardized_l_smiles).difference(set(standardized_r_smiles))) == 0
+        return len(set(standardized_l_smiles).difference(set(standardized_r_smiles))) == 0, set(
+            standardized_l_smiles
+        ) == set(standardized_r_smiles)
 
 
 class Standardizer(ABC):

utilities/ml.py

@@ -238,9 +238,11 @@ class RuleBasedDataset(Dataset):
     ):
         if feat_funcs is None:
             feat_funcs = [FormatConverter.maccs]
-        _structures = set()  # Get all the structures
+
+        _structures = set()
         for r in reactions:
             _structures.update(r.educts.all())
+
             if not educts_only:
                 _structures.update(r.products.all())
 
@@ -282,17 +284,14 @@ class RuleBasedDataset(Dataset):
                     if key not in triggered:
                         continue
 
-                    # standardize products from reactions for comparison
-                    standardized_products = []
-                    for cs in r.products.all():
-                        smi = cs.smiles
-                        try:
-                            smi = FormatConverter.standardize(smi, remove_stereo=True)
-                        except Exception:
-                            logger.debug(f"Standardizing SMILES failed for {smi}")
-                        standardized_products.append(smi)
-                    if len(set(standardized_products).difference(triggered[key])) == 0:
+                    if FormatConverter.smiles_covered_by(
+                        [prod.smiles for prod in r.products.all()],
+                        list(triggered[key]),
+                        standardize=True,
+                        canonicalize_tautomers=True,
+                    ):
                         observed.add(key)
+
         feat_columns = []
         for feat_func in feat_funcs:
             if isinstance(feat_func, Descriptor):
@@ -301,6 +300,7 @@ class RuleBasedDataset(Dataset):
                 feats = feat_func(compounds[0].smiles)
             start_i = len(feat_columns)
             feat_columns.extend([f"feature_{start_i + i}" for i, _ in enumerate(feats)])
+
         ds_columns = (
             ["structure_id"]
             + feat_columns
@@ -334,7 +334,9 @@ class RuleBasedDataset(Dataset):
                     obs.append(None)
                 else:
                     obs.append(0)
+
             rows.append([str(comp.uuid)] + feats + trig + obs)
+
         ds = RuleBasedDataset(len(applicable_rules), ds_columns, data=rows)
         return ds
 
@@ -680,9 +682,13 @@ class RelativeReasoning:
 
     def predict(self, X):
         res = np.zeros((len(X), (self.end_index + 1 - self.start_index)))
+        immutable_res = np.zeros((len(X), (self.end_index + 1 - self.start_index)))
 
         # Loop through all instances
         for inst_idx, inst in enumerate(X):
+            for i, t in enumerate(inst[self.start_index : self.end_index + 1]):
+                immutable_res[inst_idx][i] = t
+
             # Loop through all "triggered" features
             for i, t in enumerate(inst[self.start_index : self.end_index + 1]):
                 # Set label
@@ -696,7 +702,7 @@ class RelativeReasoning:
                             if i2 in self.winmap.get(i, []):
                                 # if thatat rule also triggered, it dominated the current
                                 # set label to 0
-                                if X[inst_idx][i2]:
+                                if immutable_res[inst_idx][i2]:
                                     res[inst_idx][i] = 0
 
         return res