Modeling Enzymatic Reactions in Proteins.

We will discuss application of our density functional (DFT)-based QM/MM methodology to modeling a variety of protein active sites, including methane monooxygenase, myoglobin, and cytochrome P450.~ In addition to the calculation of intermediates, transition states, and rate constants, we will discuss modeling of reactions requiring protein conformational changes. Our methodology reliably achieves small errors as a result of imposition of the QM/MM boundary.~ However, the accuracy of DFT methods can vary significantly with the type of system under study. We will discuss a novel approach to the reduction of errors in gradient corrected and hybrid DFT functionals, using empirical localized orbital corrections (DFT-LOC), which addresses this problem effectively.~ For example, the mean unsigned error in atomization energies for the G3 data set using the B3LYP-LOC model is 0.8 kcal/mole, as compared with 4.8 kcal/mole for B3LYP and 1.0 kcal/mole for G3 theory.