Study of Tamiflu sensitivity to variants of A/H5N1 virus using different force fields

An accurate estimation of binding free energy of a ligand to receptor ΔG(bind) is one of the most important problems in drug design. The success of solution of this problem is expected to depend on force fields used for modeling a ligand-receptor complex. In this paper, we consider the impact of four main force fields, AMBER99SB, CHARMM27, GROMOS96 43a1, and OPLS-AA/L, on the binding affinity of Oseltamivir carboxylate to the wild-type and Y252H, N294S, and H274Y mutants of glycoprotein neuraminidase from the pandemic A/H5N1 virus. Having used the molecular mechanic-Poisson-Boltzmann surface area method, we have shown that ΔG(bind), obtained by AMBER99SB, OPLS-AA/L, and CHARMM27, shows the high correlation with the available experimental data. They correctly capture the binding ranking Y252H → WT → N294S → H274Y observed in experiments (Collins, P. J. et al. Nature 2008, 453, 1258). In terms of absolute values of binding scores, results obtained by AMBER99SB are in the nearest range with experiments, while OPLS-AA/L, which is applied to study binding of Oseltamivir to the influenza virus for the first time, gives rather big negative values for ΔG(bind). GROMOS96 43a1 provides a lower correlation as it supports Oseltamivir to be more resistant to N294S than H274Y. Our study suggests that force fields have pronounced influence on theoretical estimations of binding free energy of a ligand to receptor. The effect of all-atom models on dynamics of the binding pocket as well as on the hydrogen-bond network between Oseltamivir and receptors is studied in detail. The hydrogen network, obtained by GROMOS, is weakest among four studied force fields.