Correlation of binding affinities with non-bonded interaction energies of thrombin-inhibitor complexes

Several empirical modeling protocols are evaluated allowing a quantification of the interaction between an enzyme and a series of inhibitors. The evaluation and optimization of the modeling protocols used a database of 35 non-covalently bound inhibitors of human thrombin. Intermolecular interaction energies were calculated with CHARMm after energy minimization of the modeled complexes using various dielectric functions and constraining strategies. These calculated binding energies were correlated with the experimentally obtained binding constants of the inhibitors. The best protocols resulted in linear correlations with correlation coefficients > 0.80. In the best protocols the enzyme was fully constrained, the ligand was allowed to move freely and electrostatics were scaled down drastically or fully neglected during the energy minimization. For the interaction energy evaluation step, a distance-dependent dielectric function epsilon = R proved to be optimal. This simple empirical protocol, that neglects solvation or entropy effects, can be implemented readily in other force field packages and may be applied on various enzyme-inhibitor complexes, providing a tool for the evaluation and rank-ordering of newly designed inhibitors.