SciDEX has 841 debate sessions and 1,878 hypotheses spanning AD, PD, ALS, FTD, and other neurodegenerative conditions — but they are siloed. No cross-disease synthesis surface exists despite the fact that shared mechanisms (TDP-43 in ALS+FTD, tau in AD+FTD+PD, neuroinflammation via TREM2 across all) are among the most actionable scientific insights the platform could produce.
Produce 10+ cross-disease mechanistic hypotheses, each with:
1. Query SciDEX for hypotheses grouped by target gene/pathway across disease fields
2. Identify genes/pathways appearing in ≥2 disease contexts with debate support
3. Search PubMed for evidence of the shared mechanism using paper_cache.search_papers()
4. For each cross-disease signal: generate a structured hypothesis with:
- Mechanistic explanation (A→B→C across diseases)
- ≥2 real PubMed citations (verify PMIDs)
- Falsifiable prediction with specific experimental test
- Proposed experiment protocol
- Cross-disease confidence score (0-1)
5. Insert hypotheses into DB with hypothesis_type='cross_disease_synthesis', disease='multi'
6. Add KG edges for cross-disease connections discovered
Starting candidates: TDP-43 (ALS/FTD/AD), tau (AD/FTD/PD), alpha-synuclein (PD/DLB/MSA), TREM2 microglial activation (AD/PD/ALS), mitochondrial dysfunction PINK1/Parkin (PD/AD/ALS), autophagy/UPS protein clearance (AD/PD/ALS/HD).
Do NOT insert hypotheses with evidence_for=null. All PMIDs must be verified real.
Spec: docs/planning/specs/quest_agora_cross_disease_synthesis_spec.md
Last Error
iteration cap hit (3/3); last verdict=needs_iteration