**Problem:** The top 25 hypotheses by composite_score (≥0.88) have been debated but never computationally validated. Forge layer tools for enrichment, expression correlation, and molecular interaction sit unused for systematic validation.
**Goal:** For each of the top 25 hypotheses by composite_score, run ≥2 computational analyses using Forge tools and update data_support_score.
**Implementation:**
1. Query top 25: `SELECT id, title, target_gene, target_pathway, disease, composite_score FROM hypotheses WHERE composite_score >= 0.88 ORDER BY composite_score DESC LIMIT 25`
2. For each hypothesis, use available Forge tools:
- Gene/pathway enrichment (Reactome pathways, enrichr-analyze skill)
- Expression context (GTEx tissue expression, Allen Brain Atlas if relevant)
- Protein-protein interactions (String DB via string-protein-interactions skill)
3. Compute data_support_score: 0.0-1.0 based on enrichment hits, expression matches, PPI confidence
4. UPDATE hypotheses SET data_support_score = , gene_expression_context = WHERE id =
5. Create an analysis artifact documenting which hypotheses gained/lost support
6. Flag hypotheses where computational evidence contradicts debate-based score (highest scientific value)
**Use `scidex.core.database.get_db()` for DB writes. Use `tools.py` wrappers or skill calls for scientific APIs.**
**Expected outcome:** 25 hypotheses with updated data_support_score. Analysis artifact created. Contradictions between computational and debate evidence explicitly documented.