Skip to content

Merge the one duplicate individual; explain why other label matches stay apart - #148

Merged
KrishnaTO merged 2 commits into
mainfrom
claude/ontology-individual-merging-262d74
Oct 16, 2026
Merged

KrishnaTO merged 2 commits into
mainfrom
claude/ontology-individual-merging-262d74

Conversation

@KrishnaTO

Copy link
Copy Markdown
Owner

Summary

Reviewed every set of individuals in ontologies/ari_t1d.owl that share a label (case-insensitive): 428 groups, 4,192 of 6,674 individuals.

Merged (1): Raynaud's (ARI:0001176) listed "Bluish coloring of hands/feet" twice (Sym_0001176_1700, Sym_0001176_1701), with no other data on either. _1701 and its hasSymptom link are removed, and the disease gets an ARI_ChangeLog line.

Not merged, and why. In this model each individual is "this symptom/antibody/cell in this disease", and it holds that disease's data.

Group Count Why it stays apart
Shared label, different content 63 groups The label is a heading and the facts sit in other properties: 1,085 pathway steps labelled "Step 1"–"Step 5", 371 epidemiology figures, and 515 "Molecular mimicry" etiologies that each name a different trigger. Values differ by disease, e.g. anti-GAD65 is 70–80% in T1D, 90–100% in LADA and under 5% in fulminant T1D. 40 of 47 affected-cell groups have a different relevance per disease.
Different classes 8 groups Epilepsy as a disease and as a symptom; TSHR antigen vs anti-TSHR antibody; MPO as a NETosis component and as an antigen; DCs as an APC and as an affected cell.
Identical today 357 groups / 1,971 individuals Merging would lose nothing now, but the editor treats each individual as owned by one disease. delete_item calls destroy_entity, so removing "Pain" from one disease would remove it from all 94. Edits write on the individual, so a likelihood or obsolete flag set for one disease would show on all of them. 6 groups also disagree on ARI_Obsolete.

Symptoms already have a shared identity in notebook/symptom-mapping/ari_symptoms.owl (one ARIS: concept per string, plus 63 synonym merges that label matching can't find). Sharing across diseases should link per-disease individuals to that concept, not collapse them.

Test plan

  • python .github/scripts/validate_mappings.py --since main: 0 errors, 0 warnings
  • OWL parses. The diff is 1 line added and 7 removed, with no line-ending churn.

🤖 Generated with Claude Code

KrishnaTO and others added 2 commits October 16, 2026 14:32
…s stay apart

Reviewed all 428 groups of individuals in ari_t1d.owl that share a label. Only
Raynaud's (ARI:0001176) held a true duplicate: "Bluish coloring of hands/feet"
twice with no other data, so Sym_0001176_1701 and its hasSymptom link are
removed and the disease gets an ARI_ChangeLog line.

The rest are per-disease records (different steps, figures, triggers and
values under a shared heading label), different classes with the same name, or
identical individuals the editor treats as owned by one disease: its delete
destroys the entity and its edits write on the individual, so a shared one
would be removed or changed for every disease at once. The changelog entry
gives the counts and reasons.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@KrishnaTO
KrishnaTO merged commit cc75a37 into main Oct 16, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant