Understanding the HIV-1 protease reactivity with DFT: what do we gain from recent functionals?

The modeling of HIV-1 plays a crucial role in the understanding of its reactivity and its interactions with specific drugs. In this work, we propose a medium sized model to test the ability of a variety of quantum chemistry approaches to provide reasonable geometric parameters and energetics for this system. Although our model is large enough to include the main polarizing groups of the active site, it is small enough to be used within full quantum studies up to the second order Møller-Plesset (MP2) level with extrapolations to coupled cluster CCSD(T) level. These high level calculations are used as reference to assess the ability of electronic structure methods (semiempirical and DFT) to provide accurate geometries and energies for the HIV-1 protease reaction. All semiempirical methods fail to describe the geometry of the protease active site. Within DFT, pure generalized gradient approximation (GGA) functionals have difficulty in reproducing the reaction energy and underestimate the barrier. Hybrid and/or meta GGA approaches do not yield a consistent improvement. The best results are obtained with hybrid GGA B3LYP or X3LYP and with hybrid meta GGA functionals with a fraction of exact exchange around 30-40%, such as M06, B1B95, or BMK functionals. On the basis of these results, we propose an accurate and computationally efficient strategy, employing quantum chemistry methods. This is applied here to study the protonation state of the reaction intermediate and could be easily used in further QM/MM studies.