CROWN: a repair-first protein-ligand dataset for training scoring models
- CROWN is a machine-learning-ready set of protein-ligand complexes from the PDB, curated by repairing structural inconsistencies wherever possible and filtering only when a repair would not be conservative.
- It is several times larger and more protein-diverse than the earlier SPINDR set, and ships with similarity annotations and a web interface.
- Retrospectively, scoring models trained on CROWN did best among five training resources on CASF-2016, and matched PDBbind- and HiQBind-trained models on docking and screening tasks.
💡 Points Based on the abstract
Monday: pull CROWN and check its similarity annotations against our own targets before retraining anything. A CASF-2016 win is weak evidence, since that benchmark is saturated and overlaps heavily with PDB training data. The parity on docking and screening is the more honest signal. Retrospective only; what matters is whether repaired structures help on our series.
📄 Paper · Journal of Chemical Information and Modeling ↗ CROWN: Curated Repository of Well-Resolved Noncovalent Interactions