Data set of protein-ligand complexes with reliable experimental structures and affinities.

The data set has been introduced in the following paper, please refer to it for more details and don't forget to cite it:
A. Pecina, J. Fanfrlík, M. Lepšík and J. Řezáč; Nature Communications 2024, 15, 1127.

The data set is also archived at Zenodo:

The data set consists of 10 target proteins, each with a series of ligands (164 in total). Crystal structures of the P-L complex are available for the majority of the ligands (147). The PL-REX set uses a single crystal structure of the protein, carefully selected so that it can accommodate all the ligands, into which the ligand poses have been superimposed. In the resulting structures, the ligand and surrounding protein residues have been optimized, yielding the final geometry presented in the data set.

The top-level numbered directories contain data for the ten protein targets. Individual P-L complexes are identified by their PDB code (majority of the systems) or labeled as a model (some ligands in 002-HIV-PR and 005-Cath-D that were modeled from similar structures).

The structures of the optimized P-L complexes are stored in the subdirectory structures_pl-rex of each protein target. Each P-L complex has its own subdirectory containing:

• ligand.sdf - The geometry of the ligand in the optimized complex.
• receptor.pdb - Model of the active site used in the complex.

These files represent the refined "PL-REX" geometry of the complexes on which the SQM2.20 score was calculated. Additionally, the optimized active site has been ported back into the structure of the whole protein and is provided as protein.pdb.

Some of the calculations reported in the paper were carried out on structures of the complexes optimized with AMBER/GAFF2 forcefield. These are available in the AMBER subdirectory of each P-L complex. There are two variants, with ligand optimized in frozen protein (complex_lig_opt.pdb) and with relaxed flexible region around the ligands (analogously to the PL-REX geometries, complex_flex_opt.pdb). In addition to the complex (the active site model with the ligand) in the PDB format, the ligand is also provided as a SDF file (ligand_lig_opt.sdf and ligand_flex_opt.sdf, respectively).

The smaller model (to which the DFT was applied) is available in the structures_small_model directories. Note that the small model has been optimized independently and therefore the geometries of the P-L complexes differ from those in the larger model.

The crystal structure of the protein from which all the P-L complexes were derived, protonated and prepared for calculations, is provided as the PDB file structures_crystal/protein_crystal_geo.pdb. In the same directory, the initial geometries of the ligands (obtained by overalapping the crystal structures oftheir complexes, protonated, but not optimized) are provided as SDF files.

The experimental affinities have been converted to binding free energies and are available in the experimental_dG.txt files.

The SQM2.20 score computed on the PL-REX geometries is available as score_SQM2.20.txt for each target. The same score, and the DFT score, compute on the small model are available in the structures_small_model subdirectory.

Scores obtained with standard scoring functions on the PL-REX geometries are available in the scores subdirectory.