Development & application of computational methods for the study of protein dynamics with PmHMGR as a model system

点击翻译按钮获取中文摘要

PROJECT SUMMARY Diseases are frequently caused by dysfunction of proteins in the body, perhaps due to maladapted genetics or from a wide variety of other causes. Researchers can gain a glimpse into this function through the study of a protein’s mechanism and dynamics. Ideally, a complete understanding of the role of a protein in biophysical interactions would describe the entire mechanistic pathway on an atomistic and dynamic level. However, this cannot be attained with experimental studies alone with today’s capabilities. Computational studies can provide experimentally inaccessible quantitative and atomistic information so they serve as powerful tools for better understanding diseases and identifying targets for experimental follow-up and potential treatment, but they carry little weight without rigorous experimental validation. We seek to reconcile experimental and computational data, equipping researchers with a method to produce the aforementioned continuous and atomistic information on protein dynamics so that they can elucidate the long timescale dynamics of proteins on an atomic level. When deconvolving time-resolved crystallographic data, I will substitute the typical static crystallographic initial inputs with structures from molecular dynamics simulations and predictive models to improve the continuity and accuracy of deconvoluted data. The objective of this work is to produce the aforementioned ideal dynamics information for a significant portion of the mechanism of PmHMGR as a demonstration and refinement of the proposed Markov State informed Multilinear Singular Value Decomposition (MSiMSVD) method which reconciles experimental and computational data. Application of the MSiMSVD method to slow dynamical events, such as the PmHMGR 2nd hydride transfer, is limited by the ability of molecular dynamics to perform accurate long-timescale simulations. This often requires Transition State Force Fields (TSFFs), but their parameterization for biomolecules often falls into local optimization minima due to high dimensionality. To reduce local minima trapping and make TSFF generation more accessible for biophysical research, I will apply constraints and swarm intelligence techniques to improve current TSFF parameterization. Collectively, these aims will provide a means by which experimental and computational techniques can work synergistically to produce the continuous atomistic protein dynamics information ideal for the investigation of proteins and their related functions and diseases.