A Computational Method Including Protein Flexibility to Estimate Affinities with Small Ligands

Abstract

Computational methods are used to generate protein-ligand complex structures and predict their binding affinities. Usually protein flexibility is not fully considered, and the energy functions used to estimate binding affinities are poor. This work investigated how implicit ligand and solvation theories as well as the linear response approximation may be combined to better describe the effect of protein flexibility on ligand binding affinities.T4 lysozyme mutants, HIV-1 reverse transcriptase (HIVRT) and human FK506 binding protein 12 (FKBP) were chosen as model systems. An adaptive energy function based on the linear interaction energy approximation was parametrized and used to estimate partial affinities. Parameters were adapted according to ligand and protein surface polarities. Proteins were represented as an approximate conformational ensemble derived from molecular dynamics simulations. Interaction energies were obtained using the OPLS-AA force field with modified partial charges for ligands. A generalized Born model was used for implicit solvation.The parametrized energy function resulted in average deviations between experimental and calculated affinities of 1.0 kcal/mol and a correlation coefficient R2=0.8 for a test set of complexes with known binding sites. Discrimination of false-positive poses was also substantial. Then, approximations to the implicit ligand theory were proposed in order to obtain total binding affinities by combining interaction energies calculated for ligand complexes with the protein conformational ensembles. Several configurations contribute with the same weight for the FKBP protein. But, for lysozyme and HIVRT proteins, total affinities are dominated by one configuration. These results suggest that a faithful representation of protein conformational flexibility and an adequate statistical treatment based on implicit theories may be used to rapidly estimate reliable binding affinities.