Stage: Hit ID
22 benchmarks, ranked by composite score. ← All stages
| # | Benchmark | Description | Modalities | Score | Flags |
|---|---|---|---|---|---|
| 1 | SAbDab | Structural antibody database — curated PDB antibody structures with annotation. | biologic-mab | 100.0 | |
| 2 | ChEMBL | Manually curated bioactive molecule DB; backbone for most ML chemistry benchmarks. | small-moleculeprotein-general | 97.5 | |
| 3 | Observed Antibody Space (OAS) | Repository of >1B antibody BCR sequences from public repertoires — core for antibody LM pretraining. | biologic-mab | 97.5 | |
| 4 | PoseBusters | Physics-aware eval of docking/co-folding poses — 19 checks catching chemically impossible outputs. | small-moleculeprotein-general | 97.0 | |
| 5 | PLINDER | Leakage-aware protein-ligand interaction dataset with stratified splits. | small-moleculeprotein-general | 97.0 | |
| 6 | CASP15 | CASP15 blind structure prediction including multimer and ligand-bound categories. | protein-general | 94.9 | |
| 7 | CASP16 | CASP16 (2024) — expanded multimer, RNA, and ligand prediction. | protein-generalsmall-molecule | 94.4 | |
| 8 | CAMEO weekly targets | Continuous weekly blind eval using pre-release PDB structures — 3D, multimer, ligand pocket. | protein-generalsmall-molecule | 94.4 | |
| 9 | canSAR | Cancer translational knowledge base integrating pharmacology, bioactivity, structure, druggability. | small-moleculeprotein-general | 89.4 | |
| 10 | PINDER | Protein-protein docking benchmark with rigorous split design. | protein-general | 88.9 | |
| 11 | CoV-AbDab | Coronavirus antibody database — binding & neutralization annotations for SARS-CoV-2/MERS/etc. | biologic-mab | 88.9 | |
| 12 | PubChem BioAssay | NIH PubChem's screening assay repository. | small-molecule | 88.6 | |
| 13 | ISM Benchmarks: GPCRs (Insilico) | 87-benchmark GPCR affinity suite on ScienceAIBench / InsilicoBench / DDB — kinome-scale coverage of class A/B/C GPCRs. | small-moleculeprotein-general | 87.6 | |
| 14 | CAPRI Rounds | 56 rounds of blind protein-protein (and peptide, ligand) complex prediction. | protein-general | 86.3 | |
| 15 | DOCKSTRING | 260k ligands docked to 58 targets with AutoDock Vina; reproducible docking benchmark. | small-molecule | 81.3 | |
| 16 | LIT-PCBA | Literature-curated dose-response PubChem BioAssay set; 15 targets with unbiased actives/inactives. | small-molecule | 80.8 | |
| 17 | MoleculeNet | Multi-task molecular ML benchmark covering quantum, physical, biophysical, physiological properties. | small-molecule | 78.0 | data-leakage-known |
| 18 | IgLM / AntiBERTa benchmarks | Antibody LM eval — paratope prediction, CDR generation, developability. | biologic-mab | 77.5 | |
| 19 | CASF-2016 | Comparative Assessment of Scoring Functions — scoring, ranking, docking, screening power tests. | small-moleculeprotein-general | 76.2 | |
| 20 | PDBbind | Curated PDB complexes with experimental binding affinities — de facto standard for ML affinity prediction. | small-moleculeprotein-general | 75.9 | data-leakage-known |
| 21 | DUD-E | Directory of Useful Decoys — Enhanced. 102 targets × actives + property-matched decoys. | small-molecule | 72.9 | data-leakage-knowndeprecated-recommend-replace |
| 22 | DEKOIS 2.0 | Decoy sets matched to actives by physicochemical properties for structure-based VS. | small-molecule | 57.5 | deprecated-recommend-replace |