The ligands herein vary in the number of atoms and number of rotatable bonds. Furthermore, most of these protein-ligand complexes were chosen because they were particularly sensitive to the performance of different search algorithms.
This is a subset of the data used in the following study: Accelerating AutoDock4 with GPUs and Gradient-Based Local Search https://dx.doi.org/10.26434/chemrxiv.9702389.v1
These structures, both proteins and ligands, were prepared in an automated way without manual inspection. The following is a non-comprehensive list of issues that may exist:
- missing water molecules that bridge ligand-receptor interactions,
- missing atoms in the proteins,
- non-integer sum of partial charges,
- incorrect protonation state,
- incorrect protein conformations.
This intended use of this data is to evaluate the performance of docking with regard to computational and algorithmic efficiency, but not the accuracy of the scoring function.
In ligand_properties.csv
pdb
Protein Data Bank accession coden_atom
number of atoms in the ligandn_tors
number of bonds in the ligand that can rotate during docking.score_of_probable_global_minimum
lowest (best) score (score is the sum of intra- and inter-moleculer energy components). The best score did not improve with an increase in the search effortRMSD_of_probable_global_minimum
RMSD from X-ray pose for the solution with the lowst scorebest_score_so_far
lowest score ever found but even lower scores are likely to exist