[KrishnaTO]

Why

PR #148 kept same-label Symptom individuals apart. Each one is a symptom in one disease: it holds that disease's likelihood, description and source. The editor also destroys an individual when it is removed from a disease, so a shared individual would be removed or edited for every disease at once. This PR keeps the individuals and links each one to its shared ARIS symptom (ARI#142). "Fatigue" across its 76 diseases can then be queried as one thing:

PREFIX ari:  <http://www.aurint.org/ontologies/ari-disease#>
PREFIX aris: <http://www.aurint.org/ontologies/ari-symptoms#>
SELECT ?disease WHERE { ?disease ari:hasSymptom ?s . ?s ari:symptomConcept aris:ARIS_0000190 }

What

• symptomConcept is a new annotation property in ontologies/ari_t1d.owl. Its value is the IRI of the ARIS_… class in notebook/symptom-mapping/ari_symptoms.owl. It is an annotation property rather than an object property because the target is a class in another file.
• The TSV owns the link. This is option A from the design discussion. symptom_mapping.tsv, keyed by symptom string, decides the concept, and build_symptom_mapping.py writes it.
  • A merged string links the class of the row it merges into. A not a symptom string links nothing.
  • The script edits only the symptomConcept lines, and a second run changes nothing.
• Coverage: 2,248 of the 2,294 Symptom individuals are linked, to 763 classes. That is every live individual (2,242) plus 6 obsolete ones whose labels map. The other 46 are obsolete not a symptom entries.
• One direction only: ari_symptoms.owl no longer carries the reverse ariSymptomIndividual links.
• Validator: validate_mappings.py checks each Symptom individual's link against the TSV, and the PR gate now also runs when the TSV changes.
  • A link that disagrees with the TSV is an error.
  • A label with no row yet, or a mapped label with no link yet, is a warning. This is the state after a curator adds or renames a symptom in the editor, and it shouldn't block their publish.
  • Each finding is placed on the label line, so a rename shows up in --since mode.
• Report 4: the Diseases (ontology) counts are corrected for 50 symptoms. The old counts still included individuals retired or merged in Retire 46 non-symptom Symptom individuals (#132) #143, Retire Long COVID symptoms copied from autoimmune pancreatitis #144, Merge the one duplicate individual; explain why other label matches stay apart #148 and Add clinical biomarkers to the data model #149, whose PRs didn't rebuild the report. Only that column changes.

Checked

• python .github/scripts/validate_mappings.py: 0 errors, 5 warnings. The 5 warnings were already on main.
• --since origin/main: 0 errors, 0 warnings.
• Three changes tried on a temporary copy, each caught as expected:
  • renamed label: symptom-unmapped warning
  • removed link: symptom-concept-missing warning
  • re-pointed link: symptom-concept-mismatch error
• The ontology diff against main is additions only: 2,253 lines, which are the 2,248 links plus the 5-line declaration.
• owlready2 0.51 loads the file, reads symptomConcept, and keeps all 2,248 links through a load and save. The editor's saves won't drop them.

Follow-up (app repo)

KrishnaTO/ARI-metadata-manager needs a change so that a curator adding or renaming a symptom sets symptomConcept from the TSV row for the new label, or clears it when there is no row. Until then, those symptoms show the warnings above until the next build.

🤖 Generated with Claude Code