CrossDocked Dataset

This dataset contains protein-ligand complexes from the CrossDocked database.

Dataset Structure

crossdocked/
├── pl/
│   ├── crossdocked_v1.1_rmsd1.0_pocket10_processed_fullatom.lmdb  # Main LMDB database (166,500 entries)
│   ├── crossdocked_v1.1_rmsd1.0_pocket10_processed_fullatom.lmdb-lock
│   ├── crossdocked_name2id.pt  # Mapping from (PDB, SDF) pairs to LMDB indices
│   └── split_by_name_10m.pt    # Train/Val/Test split (100 test samples)
└── crossdocked_test/           # Test set files (100 protein-ligand complexes)
    └── [protein_dirs]/
        ├── *.pdb  # Protein structure
        └── *.sdf  # Ligand structure

Dataset Statistics

• Total Entries: 166,500 protein-ligand complexes
• Test Set: 100 complexes
• Format: LMDB database with pickled EasyDict objects

Data Fields

Each entry in the LMDB database contains:

• protein: Protein structure data (element, positions, backbone info, atom names, etc.)
• ligand: Ligand molecule data (SMILES, element, positions, bonds, features, etc.)
• entry: Metadata list
• id: Unique identifier

Citation

If you use this dataset, please cite the original CrossDocked paper:

License

Please refer to the original CrossDocked dataset license.