Hartshorn et al. (2007), Journal of Medicinal Chemistry, Astex Diverse Set download (CCDC), PoseBusters paper data (Zenodo), RCSB Protein Data Bank and Chemical Component Dictionary
The Astex Diverse Set is a benchmark of 85 protein-ligand crystal structures published by Michael Hartshorn and colleagues at Astex Therapeutics in 2007. It was built to validate docking programs on complexes that are both high quality and relevant to drug discovery, and it remains one of the most widely used redocking tests.
This table lists each complex by PDB ID and ligand code, with the ligand structure as SMILES, the protein it binds, and the crystallographic resolution, so you can review the set or pick complexes to redock.
How the set was assembled
The authors classified publicly available crystal structures and kept high-resolution complexes whose ligand could be checked against electron density reconstructed from the deposited structure factors. Complexes were clustered by protein sequence, clusters were dropped if the protein was not thought to be of direct interest to the pharmaceutical or agrochemical industry, and rules excluded ligands that are not drug-like. One complex of sufficient quality was then chosen from each cluster.
In the original paper, the docking program GOLD redocked ligands into their native structures with success rates of roughly 80% using relatively unbiased protocols.
Why newer benchmarks exist
Most of these complexes predate the training sets of modern deep learning docking models. The PoseBusters authors found that 47 of the 85 complexes are in the PDBbind 2020 General Set and that 67 of the 85 proteins share more than 95% sequence identity with proteins in it, so strong results on this set say little about how a learned model handles new targets. The sequence identity column comes from that analysis.
The set is still useful for checking that a docking protocol is set up correctly, since classical methods are expected to redock most of these ligands. For a test on unseen complexes, use the PoseBusters Benchmark set.
Using the set in ProteinIQ
The send button on each row loads the RCSB entry and the ligand SMILES into a docking tool such as AutoDock Vina, GNINA, or DiffDock-L. The RCSB file is the full deposited entry, including the crystal ligand and waters, so remove the ligand before docking or start from the prepared files.
A redocked pose is usually counted as correct when its heavy-atom RMSD to the crystal ligand is within 2 Å. PoseBusters on ProteinIQ reports that RMSD together with checks for geometry, stereochemistry, and clashes with the protein.
How this dataset was assembled
The 85 complexes are the Astex Diverse Set rows of the PoseBusters paper data on Zenodo, which match the set published by Hartshorn and colleagues. Release date, resolution, and the protein chain nearest each ligand come from the RCSB PDB Data API, and ligand names and SMILES from the Chemical Component Dictionary. Every SMILES string parses in RDKit, which also supplies the formula and molecular weight.
This page is an index of PDB facts. The PDB ID and ligand code identify each complex in the original set, so any entry can be checked against RCSB before reuse.
License and citation
This compilation is free to download and reuse under the Creative Commons Attribution 4.0 International license. Credit ProteinIQ and the original sources when you share it or publish results that use it.
Suggested citation: ProteinIQ (2026). Astex Diverse Set, version 2026.10.05. https://proteiniq.io/datasets/astex-diverse-set
The prepared Astex Diverse Set files are distributed by the CCDC under its own download agreement, so ProteinIQ does not host them. Get them from the CCDC downloads page or the PoseBusters Zenodo record.
Astex Diverse Set download (CCDC): Official prepared protein and ligand files. Cited and linked, not mirrored. Downloads require accepting the CCDC test set agreement.
PoseBusters paper data (Zenodo): Complex identifiers, sequence identity to PDBbind 2020, and cofactor flags. Released under CC BY 4.0.
RCSB Protein Data Bank and Chemical Component Dictionary: Release dates, resolution, the protein chain nearest each ligand with its UniProt accession, and ligand names and SMILES from the Chemical Component Dictionary. PDB data are released under CC0 1.0.
Frequently asked questions
The set contains 85 protein-ligand complexes, each with a single ligand of interest.
The prepared files are distributed by the CCDC from its downloads page after accepting its test set agreement. The PoseBusters Zenodo record also includes processed versions of all 85 complexes. ProteinIQ hosts only this index.
It is a good sanity check for a docking protocol, but most of its complexes overlap with the training data of deep learning methods. Use a time-split set such as PoseBusters to judge how a method handles new targets.
Cite Hartshorn et al., "Diverse, high-quality test set for the validation of protein-ligand docking performance", Journal of Medicinal Chemistry 50, 726-741 (2007), doi:10.1021/jm061277y.