UC San Diego

Cellular membranes are incredibly complex structures composed of diverse biomolecules including lipids, proteins, and other molecules. Cellular membranes are important because they allow for the transfer of molecules and chemical signals in and out of cells. The structure of membranes and membrane proteins is difficult to determine through experiments. Membrane structure is also highly dynamic and depends on the mixture of molecular components. While the membrane is usually in the liquid- disordered phase, other local membrane assemblies have been observed. It is challenging to predict the permeation of pharmaceutical compounds through the membrane. Molecular dynamics (MD) is a computational method that allows for the study of membrane motions and drug interactions. Based on a chemical model of molecules and Newton's equations of motion, it is possible to predict the dynamics of molecules on a computer. However, in order to simulate new molecules, it is necessary to refine an appropriate force field that models the chemical interactions of the molecules. A new force field was developed for lipids, an essential membrane component in the Amber MD software package. This force field was parameterized with experimental data and quantum mechanical calculations on individual chemical components of the lipid molecule. Afterwords, parameters were validated against available membrane structural data. Parameters have been developed for a set of glycerophospholipids and cholesterol. Molecular dynamics simulations of lipid membranes with the new parameters have accurately predicted membrane structural properties. With an accurate model of lipid membranes, it is now possible to examine complex membrane dynamics. Permeation of small molecules across the membrane is especially interesting in the pharmaceutical industry. Using the inhomogenous solubility-diffusion model it is possible to predict small-molecule permeability across a membrane from potential of mean force calculations. A constrained molecular dynamics algorithm was implemented in Amber for this task. The constraint implementation may be optimized to run on graphics processing units (GPU). This dissertation marks the first expanded lipid force field in Amber for accurate membrane simulations. It also marks the implementation of a accelerated general constraint methods in Amber