A non-equilibrium work study of antimicrobial peptide-membrane interactions
As antibiotic resistance develops among strains of infectious bacteria, a global need is growing for new antimicrobial agents. We rely today mostly on antibiotics that have been in wide use for decades, and these are becoming increasingly inefficient against several strains of multi-drug resistant bacteria. Traditional antibiotics have typically been extracted or derived from natural organic compounds, and have then been tested by standard biochemical and pathobiological methods. Many detailed molecular aspects of the antimicrobial activity are fundamentally inaccessible to such macroscopic experiments. The availability of computational power and biophysical knowledge at the molecular level enables a bottom-up approach to drug discovery, design and testing, the latter of which is at the main focus of this work. In this thesis a series of techniques have been applied and a few developed, all based on classical all-atom molecular dynamics simulations, for studying a computationally discovered antimicrobial peptide and its interactions with model phospholipid membranes. The structure of the peptide is studied through microsecond-long simulations, and the findings are compared with results from circular dichroism experiments for this molecule. Toward calculating potentials of mean force (PMFs) for interaction of this peptide with different model membranes, a comprehensive method has been developed, based on the newly discovered non-equilibrium work theorems. Pioneered by C. Jarzynski's work in 1997, these theorems aim at extracting the free energy difference between two macrostates of a system, from non-equilibrium external work performed on it over different phase-space trajectories between the two states. Proper implementation of such theorems to obtain fast convergence to the underlying PMF has been a challenge taken up by many investigators. The methods presented here are based on the forward-reverse (FR) method by I. Kosztin and coworkers, and deliver very good convergence for relatively short simulations. The design of these methods is based on simple observations in statistical physics, and their implementation relies on a new technique for post-simulation statistical mechanics analysis of the recorded external force values. As a result of this study, a detailed procedure is provided, with recommendations on good practices and warnings about sources of systematic errors. The presented methods are used to calculate the PMF profiles for the peptide-membrane systems under study, as well as for two other test systems. The obtained peptide-membrane PMFs are used to extract free energies of binding and their entropy-enthalpy decomposition, and the results are compared with those from isothermal titration calorimetry experiments on the same systems. Using the series of methods presented, and with availability of proper computational resources, the whole simulation procedure and extraction of results for similar systems carried out by us should be performable in a matter of a few weeks, which can speed up the drug-testing procedures, as well as any other lines of study that requires calculating PMF profiles for molecular systems.