A standalone structure-based screen of an 8.4M in-stock library against 865 curated protein targets. Each target's ~10k target-tailored shortlist was docked (Uni-Dock) and rescored (gnina CNNaffinity, or Vina for nuclear receptors) — pick a target, read its ranked hits, and click any one to see the docked pose in the pocket. Scores are ranking scores, not calibrated pIC50.
| # | CNNaff | Vina | SMILES | |
|---|---|---|---|---|
| Select a target to load its ranked hits… | ||||
How to read this: hits are ranked by gnina CNNaffinity (higher = stronger predicted binding; the pose-aware CNN rescore) or, for nuclear receptors, by Uni-Dock Vina (more negative = better). These are relative ranking scores for triage — not calibrated affinities. Shortlists are diversity-selected library analogues in the target's chemical space, not known actives; treat as leads to test. Physically clashing poses (Vina > 0) and PAINS (orphan targets) are filtered out; 12 targets with pockets too small for drug-like matter are excluded. Click → design on any hit to carry it into the Drug Design tool for prediction, neighbours and matched-pair analysis.