Affordable Access

Decompositions of Free Energies in Molecular Simulation

  • Irwin, Benedict William John
Publication Date
Sep 14, 2018
Apollo - University of Cambridge Repository
External links


This thesis describes advances in methods to measure free energy changes in simulations of molecular systems. In each case the free energy is decomposed into local environments which reveal insights about the complex systems being studied. Free energy is a fundamental quantity that can be used to predict whether changes in state are physically favourable. This can be used to predict the solubility of molecules and whether molecules are likely to bind to proteins. There are a handful of methods which measure free energy from molecular simulations. In chapter 3 we show results for an improved endpoint free energy method using inhomogeneous fluid solvation theory (IFST) which takes second order fluid-fluid entropy corrections into account. This is applied to a system of Lennard-Jones particles which show no measurable second order entropy contribution which fits with theoretical predictions. In chapter 4 an adaptation to the Zwanzig equation for path based exponential averaging methods is made. The equation is expanded to give contributions associated with every atom in the system. This method is called atomwise free energy perturbation and is applied to small molecules and ligand-protein binding. In chapter 5, IFST is applied to decompose hydration free energy at the surface of a protein into hydration sites. From these sites, information is inferred about the binding conformation of two proteins GABARAP and the GABA-A receptor. In chapter 6 statistics from hydration sites around hundreds of proteins are analysed. The distributions of free energy are shown and discussed for hydration sites in a range of local chemical environments. Also in chapter 6, the hydration sites decomposition method is augmented with local energy information associated with replacing a water molecule at a hydration site with a probe. The probe represents a ligand, and this is compared to the binding site prediction from the previous method. Further suggestions for improvements are made.

Report this publication


Seen <100 times