SciDEX has 77 hypotheses with composite_score ≥ 0.8 and 1,878 total hypotheses, but these exist as database rows with partial scores and fragmentary evidence. They lack structured scientific output that would make them actionable to researchers, funders, or drug discovery teams.
Generate structured research briefs for the top 25 hypotheses (by composite_score, minimum 0.7):
Each brief (~1,500-2,500 words) must cover:
1. Executive summary (3-4 sentences): the core claim and neurodegeneration relevance
2. Mechanistic model: A→B→C causal chain with evidence
3. Evidence synthesis: evidence_for/evidence_against with PMID evaluation
4. Debate synthesis: summary of debate sessions linked to this hypothesis's analysis_id
5. Falsifiable predictions: 3-5 specific testable predictions with experimental designs
6. Therapeutic angles: druggable targets, existing compounds
7. Confidence assessment: explanation of the 8-dimensional score
8. Open questions: unresolved aspects (feeds back to knowledge_gaps)
9. Suggested next experiments: ranked by cost/feasibility/impact
For each hypothesis:
- Collect evidence entries, look up PMIDs via paper_cache.get_paper()
- Find debate sessions via analysis_id
- Query KG edges involving target_gene/pathway
- Search PubMed for recent evidence via paper_cache.search_papers()
- Generate brief, store as 'analyses' artifact with analysis_type='research_brief'
- Update hypothesis record with pointer to brief
All PMIDs must be real and verified. Briefs shorter than 1,000 words are incomplete.
Spec: docs/planning/specs/quest_agora_hypothesis_research_briefs_spec.md