DockBind: a quality-controlled dataset of modelled protein–ligand complexes associated with affinity data

DockBind is a curated dataset designed to support the development and benchmarking of machine learning models for protein–ligand binding prediction. It includes high-quality protein-ligand complexes modeled through molecular docking annotated with binding affinity data from ChEMBL and BindingDB. Here you can find The extended Astex set that was used to validate docking (docking_validation_data.tar.gz). This archive contains a list with all the PDB structures that were used for validation (data/astex_selected_for_docking.csv). We also relase all poses that were generated in the docking process together with their: FlexX scores (docking_results/free_docking/docked and docking_results/template_based_docking/docked) Hyde scores (docking_results/free_docking/scored and docking_results/template_based_docking/scored) MCS RMSD filtering (docking_results/free_docking/filtered_[cutoff] and docking_results/template_based_docking/filtered_[cutoff]) All docking poses that were generated when compiling DockBind (DockBind_raw_data.tar.gz), with the same data structure as mentioned above The DockBind dataset, consisting of a CSV file containing the experimental binding affinity values and a hierarchical directory structure containing the structural data associated with each protein-ligand complex. Hierarchical directory structure: complexes/├── <PDB>_<LIG>/│ ├── <PDB>.pdb│ ├── reference_ligand.sdf│ ├── <ID>/│ │ ├── <PDB>.pdb│ │ ├── <ID>.sdf│ │ └── <PDB>_<ID>.pdb└── ... Top-level directories are named as <PDB>_<LIG>, where <PDB> is the protein template PDB identifier and <LIG> is the three-character code of the crystallographic ligand. Each template directory contains the original protein structure (<PDB>.pdb), the reference ligand extracted from the crystal structure (reference_ligand.sdf), and one subdirectory per docked compound. Compound directories are named according to their source database identifier (<ID>) e.g. CHEMBLxxxxx or BINDINGDByyyyy and contain the prepared protein structure used for docking (<PDB>.pdb), the docked ligand pose (<ID>.sdf), and the corresponding protein-ligand complex (<PDB>_<ID>.pdb).