Conformational Search with Implicit Solvation Model for Prediction of Ligand-Bound Conformations

Accurate prediction of ligand-bound conformations in protein-ligand complexes only from ligand information is expected to be useful in drug design. In this study, we compared the conformations obtained by a conformational search algorithm and the geometry optimization based on quantum chemical calculations with the ligand-bound conformations obtained by the constrained geometry optimization. Conformational searches for ligands of 25 atoms or less in the Platinum Diverse Dataset were performed using the reservoir algorithm and MMFF94s force field. The geometries of the obtained conformations were optimized using quantum chemical calculations at the level of ωB97X-D/6-31+G (d,p). The ligand-bound conformations were obtained by the geometry optimization at ωB97X-D/6-31+G(d,p) with fixed dihedral angles. The environment around the ligand was considered by implicit solvation models, and better results were obtained when aqueous solvents were specified than in the case of no solvent effects or octanol solvent. The RMSD means of the most similar conformation and the ligand-bound conformation were 0.534, 0.449, and 0.427 Å for no solvent effect, octanol, and water, respectively. The result indicates the importance of incorporating the effect of the surrounding environment in the prediction of the ligand-bound conformation.