Advances in computational solvation thermodynamics

The aim of this thesis is to develop improved methods for calculating the free energy, entropy and enthalpy of solvation from molecular simulations. Solvation thermodynamics of model compounds provides quantitative measurements used to analyze the stability of protein conformations in aqueous milieus. Solvation free energies govern the favorability of the solvation process, while entropy and enthalpy decompositions give insight into the molecular mechanisms by which the process occurs. Computationally, a coupling parameter λ modulates solute-solvent interactions to simulate an insertion process, and multiple lengthy simulations at a fixed λ value are typically required for free energy calculations to converge; entropy and enthalpy decompositions generally take 10-100 times longer. This thesis presents three advances which accelerate the convergence of such calculations: (1) Development of entropy and enthalpy estimators which combine data from multiple simulations; (2) Optimization of λ schedules, or the set of parameter values associated with each simulation; (3) Validation of Hamiltonian replica exchange, a technique which swaps λ values between two otherwise independent simulations. Taken together, these techniques promise to increase the accuracy and precision of free energy, entropy and enthalpy calculations. Improved estimates, in turn, can be used to investigate the validity and limits of existing solvation models and refine force field parameters, with the goal of understanding better the collapse transition and aggregation behavior of polypeptides