The causal entity resolution task (12c461ae) raised kg_edges from 2,366 to 7,442 by resolving causal_edges entries where entity names matched existing kg_entities. But 19,753 causal_edges minus 7,442 resolved = ~12,311 edges remain unresolved because their source/target entities don't match any existing kg_entity. These unresolved edges contain the richest mechanistic knowledge in the system — debate-extracted causal claims with mechanism_description and evidence_pmids — but none of it is queryable via the KG.
**Why this matters**: The unresolved edges likely contain: (1) novel entities not yet in the KG (new protein complexes, pathways, cell states), (2) cross-disease mechanisms appearing in multiple debates (AD and PD sharing the same upstream mechanism), (3) entities described in non-canonical ways that could be normalized. Mining these creates new KG nodes that unlock the remaining 60% of causal knowledge.
**What success looks like**:
- Analysis report: top 50 most-cited unresolved entity names in causal_edges source/target
- ≥20 new KG entity nodes created for the highest-value unresolved mechanisms
- ≥500 additional causal_edges resolved into kg_edges after new entity creation
- Cross-disease pathway map: identify mechanism names appearing in debates for ≥2 of AD/PD/ALS/FTD
- Report committed as analysis artifact
**What to read first**:
- `docs/planning/specs/quest_atlas_causal_kg_entity_resolution.md` — base spec
- `causal_edges` schema: source_text, target_text, mechanism_description, evidence_pmids, confidence
- `kg_entities` schema: for understanding what entity types exist
- SQL to find top unresolved: `SELECT source_text, COUNT(*) FROM causal_edges WHERE id NOT IN (SELECT causal_edge_id FROM kg_edges WHERE causal_edge_id IS NOT NULL) GROUP BY source_text ORDER BY COUNT(*) DESC`
**What NOT to do**: Do not create KG entities for entities that appear only once (likely noise). Do not modify causal_edges — only create new kg_entities and resolve existing edges.
Completion Notes
Released by supervisor slot 12 because credential acquisition failed after pre-claim. Reason: worktree_creation_failed:branch_held_by_other:held_by=/tmp/wt-52160795
Last Error
acquire_fail:worktree_creation_failed:branch_held_by_other:held_by=/tmp/wt-52160795
Payload JSON
{
"completion_shas": [
"8aa4477ac",
"da775831b8d54858d558333b0ed7dc263d6ea80c"
],
"completion_shas_checked_at": ""
}