Ab initio methodologies in studying enzymatic reactions

UnrestrictedModeling chemical reactions in condensed phases, especially in biological environment, such as in enzymes, is a major task of computational chemistry. At the current computer resources one can usually not afford to solve the Schrödinger equation for the molecules, and not even for all the electrons of the system. Simplified classical physical models are used for many of the processes. For chemical reactions, however these classical models are insufficient and quantum treatment is necessary for the atoms participating in the reactions.; Our goal is to provide computational methods that can quantitatively reproduce experimentally observed reaction rates and free energy changes, such as activation free energies. For this purpose we developed methods that describe the reactive part of the system with ab initio calculations, providing accurate quantum mechanical treatment, and at the same time account for the environmental effects in a consistent way. My contribution by this PhD work is to offer several alternatives and examples how reactions in complex systems can be studied.; The concept behind solvation models can be looked at as the first version of "coarse graining": simplifying the fast degrees of freedom and keeping the slow ones that contribute to the high energy transition states.; Main results of the dissertation include: (1) major developments for the Langevin dipoles solvation method: new implementation in the Gaussian program package, derivation and implementation of analytical gradients, new formulation of the model that allows a variational treatment of the wave function; (2) development and application of a new all atom solvation model and implementation in the Q-Chem program package; (3) theoretical investigation of the concept of Linear Free Energy Relationships (LFERs) in the context of the transferability of the off-diagonal elements between different phases for the Empirical Valence Bond (EVB) method and in applications for phosphate hydrolysis reactions in water and in proteins; (4) development and applications of a method using reference potential energy surfaces to allow for better sampling and accuracy using Quantum Mechanics/Molecular Mechanics (QM/MM).; It is found in our work that ab initio methodologies are very powerful in accurately modeling potential energy surfaces. The usual total energy minimization is essential and it works well in most cases, however it is often difficult, especially in larger systems, to include all the degrees of freedom to follow a reaction path. In these cases one can easily end up in local minima. In other words, not having adequate sampling of all degrees of freedom to find better minima leads to artificially too high reaction barriers. We could overcome this problem by reducing the dimensionality of the problem: introducing simplified solvation models is shown to reproduce experimental activation energies for a series of reacting molecules. Alternatively, we also used reference potentials, where much better sampling could be achieved and we obtained converging results for reactions in protein and water, reproducing experimental findings