Hypothesis Prioritization
HYPOTHESISPhase 2: Multi-criteria ranked hypotheses scored across evidence depth, source convergence, therapeutic clarity, target strength, and novelty. Tier A = top 5 high-conviction, Tier B = medium priority, Tier C = needs more evidence. These are machine-generated hypotheses ranked by a computational scoring model. None has been experimentally validated; the SMN1/SMN2 entries reflect established external literature, the rest are predictions to be tested.
▶How does Hypothesis Scoring work?
Hypotheses are generated by Claude Sonnet + Gemini from clusters of co-occurring claims, then scored across 5 dimensions:
- Evidence Depth (25%) — number of independent PubMed sources supporting the core claim; weighted by study type (RCT > cohort > in-vitro)
- Source Convergence (20%) — agreement across distinct research groups; single-lab replications score lower
- Therapeutic Clarity (20%) — specificity of the proposed intervention: named target, mechanism, and disease stage
- Target Strength (20%) — composite target score of the primary molecular target (see Target Scoring methodology)
- Novelty (15%) — inverse frequency of the hypothesis pattern in the existing literature; penalizes well-known mechanisms
Final score = weighted sum. Hypotheses below 0.3 are suppressed from the default view. Generation runs nightly on newly ingested claims.
Loading hypotheses...