Computational Prediction of Ligand-Binding Sites of Proteins by Combining Hydrophobic Area Search and Residue-Based Scoring

Large-scale computational analysis enhances the efficiency and productivity of structure-based drug design (SBDD), driven by advances in computer systems. Docking simulation is a core technique in SBDD. An appropriately defined search area is essential for accurately predicting binding poses in docking simulations of ligands or compounds to target proteins. Therefore, computational searches for possible binding sites are useful for reliable simulations. In this work, ligand-binding site prediction was performed by combining a hydrophobic potential-based search and scoring using a residue-propensity-based function. First, multiple potent binding sites for a target protein were identified by hydrophobicity using grids generated around the protein. Next, the identified multiple binding sites were ranked according to the frequency of amino acid residues on the protein surface. In total, approximately five thousand protein–ligand complexes were tested in this approach. The top-ranked site matched the ligand-binding sites observed in the crystal structures in a quarter of the complexes. Although the top-ranked site did not always coincide with the experimentally observed ligand-binding area, the ranking adequately prioritized the candidate sites. In more than half of the complexes, the experimentally observed binding site was found in the top three candidates. Even when the experimentally observed ligand-binding area did not rank first, it was often among the higher-ranked sites. Therefore, the approach proposed in this work effectively generates plausible binding sites for docking simulations, thereby enhancing the accuracy of in silico compound screening in SBDD.