Estimates of relative binding free energies for HIV protease inhibitors using different levels of approximations.

Although the free energy perturbation approach is a rigorous method for estimating the relative binding free energy between an enzyme and its inhibitors, it is computationally expensive. This paper examines the accuracy at different levels of approximations, following the series expansion of free energy derived by Aqvist et al. Level-0 calculates only the enzyme-inhibitor interaction energy at the minimum energy configuration without solvent. In Level-0MD, the inhibitor configurations are sampled by molecular dynamics. These levels assume that the second- and higher order terms in the series expansion can be neglected and that the interaction energies in the bound and unbound states are equal. Level-1 does not assume equal interaction energies in the bound and unbound states. Level-1S includes the solvent contribution but both enzyme and inhibitor are fixed. In Level-1SMD, the inhibitor configurations are sampled by molecular dynamics. Level-2SMD retains the second-order term. We chose seven HIV-1 protease inhibitors for study: A77003, A76889, A76928, A78791, A74704, JG365 and MVT101. Level-0 and Level-0MD were found to give essentially the same relative interaction energies by using the AMBER force field, suggesting that fixing atomic positions may be a good approximation in some cases. However, as expected, Level-0 or Level-0MD gave poor predictions for the relative binding free energies between hydrophobic inhibitors (e.g. A77003) and more hydrophilic inhibitors (e.g. JG365). Level-1SMD produced a much better correlation between calculated and experimental results. Inclusion of the second-order term did not improve the accuracy.