[Atlas] Entity canonicalization and ontology alignment done

[Atlas] Entity canonicalization and ontology alignment

Goal

Build kg_normalize.py: (a) map relation types to controlled vocabulary of ~20 canonical relations, (b) merge duplicate entities, (c) align entity types to fixed ontology (gene, protein, pathway, disease, drug, phenotype, cell_type, brain_region). Runs in post_process.py pipeline. Acceptance: relations <=25 types; no duplicates.

Acceptance Criteria

Approach

Read relevant source files to understand current state

Plan implementation based on existing architecture

Implement changes

Test affected pages with curl

Commit with descriptive message and push

Work Log

• 2026-04-01 10:00 - Started implementation. Examining current pipeline and data structures.
• 2026-04-01 10:30 - Completed implementation:

* Built kg_normalize.py with canonicalization logic
* Mapped 77 relation types → 10 canonical relations (✓ <=25)
* Mapped 33 entity types → 5 canonical types (✓ <=8)
* Implemented duplicate entity merging (removed 405 duplicates)
* Integrated into post_process.py pipeline (step 2)
* Tested successfully: all acceptance criteria met
* Canonical relations: activates, causes, component_of, decreases_risk, encodes, increases_risk, inhibits, participates_in, produces, regulates
* Canonical entity types: disease, drug, gene, phenotype, protein

• 2026-04-25 21:15 - Rebuilt kg_normalize.py on current main:

* Canonical relations: 25 (limit 25) ✓
* Canonical entity types: 8 (limit 8) ✓
* DB had 892 distinct relation values → 25 canonical relations used ✓
* DB had 134 distinct entity types → 8 canonical types used ✓
* Normalized 104,984 relation values, 49,251 entity types
* Merged 27,422 duplicate edges on UniqueViolation
* Commit: 6253aa178 [Atlas] Add kg_normalize.py [task:ed421469-317a-46da-b33f-d440a705cd9f]