Computer simulations of molecular systems are a vital component of modern chemistry and physics. They are being used, for example, in such diverse areas of research as the fundamental physics of crystal nucleation, through to the design of new pharmaceutical entities. Research in the Essex group focuses on innovation in the application of computer simulations to biological systems, where there is the potential to contribute to drug discovery and the development of medical diagnostics. A key challenge restricting success has been limitations in the range of applicability of these computational methods, and in the extent to which accurate predictions may be made. To address these issues, a cross-disciplinary approach that develops new methodologies and deploys these over more realistic systems is used.
Conventional Monte Carlo or molecular dynamics simulations are limited in terms of their ability to sample efficiently. Even with the aid of advanced high performance computers or GPUs, simulations are limited to millisecond lengths. Slow or rare events, such as protein conformational changes, which may be crucial to biological function, may be only inadequately sampled on this timescale, or indeed not at all. To address this problem we have a long-standing interest in using and developing enhanced sampling approaches.As an example, we are working to combine grand canonical Monte Carlo methods with molecular dynamics and free energy calculations, to improve the sampling of bound water molecules in protein-ligand systems. Using this approach, water binding sites are quickly and accurately located, and adapt smoothly to changes in the chemistry of the bound ligand.
The calculation of Gibbs energies is crucially important to determine the stability of molecular systems. For example, to rationally improve the binding affinity of a potential drug, we seek to optimise the intermolecular interactions and thereby decrease the associated Gibbs binding energy. We have been working in the field of Gibbs energy calculations for many years, developing and applying new methodologies. For example, we were the first to apply replica exchange approaches to improve the precision of ligand binding free energy calculations.Part of our current work in this area involves the automation of these simulations to allow large-scale sensitivity studies to be performed. We have recently shown, for example, that the choice of initial crystal structure can have a significant effect on the calculated ligand binding free energy.
The lipid membrane is an integral part of biological systems. It is more than a passive envelope however, separating the cell contents from their environment. It also serves as a matrix in which crucial proteins are embedded, and it is known that the lipid composition can affect protein structure and function. We performed some of the earliest atomistic simulations of membrane systems, and have since developed coarse-grain simulation models of membranes, where groups of atoms are subsumed into single interaction sites. What distinguishes these models from alternatives, is that the electrostatic interactions of the lipids with their environment are more realistically captured, leading to improved physical models.Our current work in this area includes studies of peptide permeation through membranes, and the detailed study of the effect of cholesterol on the physical properties of lipid bilayers. Cholesterol is present in many biological membranes, and it can have a profound effect on the bilayer biophysics, affecting phase behaviour and elastic properties. Our simulations are seeking to understand how well molecular simulations are able to reproduce these effects.
In addition to developing new simulation approaches to study biological systems, we also seek to apply these approaches to real biological problems, often in collaboration with industrial or experimental partners. For example, we have long-standing collaborations with Cancer Immunology at Southampton General Hospital, looking to understand how molecular structure affects biological response. By combining low-resolution small-angle X-ray scattering data with high-resolution molecular simulations, we have been able to understand, at the atomic level, why certain antibodies reduce an immunological response, while others enhance it. Through our collaborations with industry, we have been applying our methods to problems of direct relevance to the drug development pipeline, seeking to improve their efficiency and accuracy.
Despite significant research in enhancing simulation sampling and developing better force fields, simulations of large scale systems is, in many cases, still not predictive. A solution to this problem is to incorporate experimental data directly into the simulation either by reweighting simulation trajectories to better reflect experimental data or by using experimental data to bias simulations as they are performed. We are working to combine both low and high resolution experimental data in the modelling of biological systems including RNA, antibodies and proteins more broadly.