All Predictions
| Rank | Disease | Drug | Score | Confidence | Novelty | Pillars | MR | Evidence |
|---|
Cross-Disease Drug Network
Drugs predicted as novel candidates for multiple diseases — cross-disease convergence from independent AI pillars suggests shared biological mechanisms and increases prediction confidence.
Top Breakthrough Predictions
Highest-scoring predictions with no prior published evidence — genuinely novel computational discoveries. Click any card for full evidence.
Methodology: 8-Pillar AI Scoring
TransE Embeddings
Knowledge graph link prediction using DRKG (5.87M biological relationships)
RotatE / PyKEEN
Complex relation pattern scoring via rotation-based graph embeddings
TxGNN
Graph neural network for therapeutic use prediction (Harvard)
Molecular Fingerprints
RDKit ECFP similarity to known treatments for each disease
ChemBERTa
Transformer-learned molecular representations capturing deep structural patterns
Gene Signatures
L1000CDS2 transcriptomic reversal — drugs that reverse disease gene expression
Network Proximity
STRING protein-protein interaction shortest paths between drug targets and disease genes
Mendelian Randomization
Causal genetic evidence from Open Targets — strongest form of drug target validation
Each drug receives a dynamic weighted score across all applicable pillars, with convergence bonuses when multiple independent methods agree. Candidates are then enriched with evidence from PubMed, ClinicalTrials.gov, FDA FAERS, Semantic Scholar, and L1000CDS2.