Base score: the implemented formula
Core: designated centre +22; scaffold +18 / payload +16 / linker +10. SMILES +12 (absent −8), InChIKey +4, seed source +3, inherited source −5. Family size min(14, round(log₂(n+1)×3.5)); edges min(8, 2×edges); derived children min(10, 3×children); notes +3.
Derivative: drug +18 / adc +16 / derivative +12; SMILES +12 (absent −8), inherited −5; centre Tanimoto round(22×value); family size min(10, round(log₂(n+1)×2.5)); edges min(8, 2×edges); notes +4. Scores are capped at 100.
Approval, clinical, lead and “hot field” tag bonuses are not used. Tags do not independently verify clinical evidence.
Bins and suggestion limits
Within a pool, approximately the top 12% receive X-S, through 32% X-A, through 62% X-B, and the remainder X-C. Pools smaller than five receive X-B. These are relative ranks within the current records, not externally calibrated grades.
Insight suggestions aggregate family counts, graph density and exploration signals. They do not assess experimental validation, intellectual property, synthetic feasibility or development prospects. Catalog coverage creates selection bias.