The B-factor index for the binding site (BFIbs) to prioritize crystal protein structures for docking

Abstract

In cheminformatics, protein-ligand docking is a powerful tool applied for virtual screening, pose prediction, and binding affinity estimation. However, docking results depend on the quality of the crystal protein structures. Although several crystallographic indices can assist the selection of proteins, these are often misused by non-experts or have little applicability. Hereby, we propose the B-factor index for the binding site (BFIbs), which indicates the atomic fluctuations of the atoms in the binding site in comparison with atoms in the protein. Using an automated docking workflow, we performed docking experiments on 26,019 protein-ligand complexes. The docking performances were analyzed based on five crystallographic quality indices, i.e., BFIbs, DPI (the diffraction–component precision index), DPIbs (DPI of the binding site atoms), RRfree (R–Rfree), and the resolution of the X-ray crystal structure. Only BFIbs was found to significantly correlate with the root–mean–square deviation (RMSD) computed between the ligand poses in the crystal structures and the predicted docking poses. The majority of the best docking results (RMSD < 2 A) were indicated by BFIbs < 1. We conclude that BFIbs, as a simple and interpretable parameter that complements other indices, can help to effectively prioritize protein structures for structure-based cheminformatics.