Assessment of solvation effects on calculated binding affinity differences: trypsin inhibition by flavonoids as a...

On the basis of molecular models of the interaction between trypsin and a series of seven structurally congeneric bioflavonoid inhibitors, the influence of solvation effects in the calculation of binding free energy differences in congeneric series has been assessed. The models were derived by making use of the X-ray crystal structure of bovine trypsin and the DOCK program, and the complementarity of the interactions between the functional groups of the docked molecules and the binding site region was corroborated independently with the GRID program. Interaction energies calculated for the complexes using molecular mechanics were found to correlate with the experimental inhibitory activities, although the quality of the correlation was dependent on the molecular modeling protocol. To understand why such correlations could be obtained in the absence of an explicit description of solvent effects, the in vitro activities were transformed into binding free energies, and continuum electrostatic theory was used to incorporate solvent effects by approximating them to the electrostatic contribution to the binding free energies. The results of our calculations show that, within this congeneric series, the cost in electrostatic free energy of desolvating both the enzyme binding site and the buried part of the inhibitors (delta Gdesolv) is roughly constant within the series. On the other hand, the electrostatic interaction energy (EeleLR) varies only slightly along the series in comparison with the van der Waals interaction (EVDWLR), and this variation is mostly solvent-independent, i.e., the reaction field energy of the solvent in the bound state (EsrfLR) makes almost a negligible contribution to the binding free energy differences. As a result, differences in binding free energy are dominated by the van der Waals term, while the electrostatic contribution is, to a good approximation, solvent-independent. A similar scenario may account for the good correlations frequently found between ligand activities and ligand-receptor interaction energies derived using plain molecular mechanics, although generality remains to be determined.