Rapid Solvation Thermodynamics Calculations for Proteins

德克萨斯大学医学分部（UTMB）的Montgomery Pettitt获得了化学学部生命过程化学项目的奖励，他创造了一套新的计算工具，用于模拟和理解溶液中蛋白质的基本特性，并应用于蛋白质设计和新材料和生物技术的开发。生物技术应用中一些最具挑战性的系统是那些涉及液体的系统，特别是生物聚合物在水中的混合物。Pettitt教授正在开发一种基于当前液体行为理论的蛋白质分子设计的新计算框架。作为结构成分和催化化学作用的蛋白质溶液是许多生物技术的核心。与现有方法相比，该研究将使提出的蛋白质修饰的计算快速筛选具有大大提高的准确性。在这个项目中开发的代码、脚本和文档可以免费提供给模拟社区，并且可以与广泛使用的分子动力学模拟代码相连接。该项目每年招收几名学生，在佩蒂特教授的实验室进行为期10周的计算生物学实习，通过当地HBCU和西班牙裔服务机构以及ACS SEED和R. a . Welch暑期学者高中项目的外展努力。与UTMB西利结构生物学中心密切合作，该中心为学生在研究专题讨论会和会议上的研究报告提供了场所，并为使用新方法的研究合作提供了机会，并为未来的改进提供了反馈。控制蛋白质、配体之间识别的机制以及蛋白质从未折叠状态到天然状态的转变仍然是基本的化学问题。这些平衡是由溶液中的自由能差决定的。通过显式计算机模拟计算溶剂化自由能可以是准确的，但是计算成本很高，并且对于整个蛋白质来说基本上是禁止的，即使在大型超级计算系统上也是如此。该项目正在开发从基准全原子分子动力学模拟中提取的预先计算溶剂分布（“近端（径向）分布函数”，或pdf），以产生任意蛋白质溶质的溶剂化密度曲线的近邻近似。pdf的初始库正在为广泛使用的CHARMM力场开发。本研究的主要目标是：(1)为所有天然存在的氨基酸和其他重要官能团生成pDF库；(2)开发基于pDF的有效方法来计算总和相对溶剂化自由能；(3)将该方法扩展到多组分溶液。第三个目标包括实验合作，以测试溶液热力学测量的方法。鉴于pdf的预先计算性质，将热力学平均值作为预先计算分布的积分的评估在计算上是非常有效的。这些快速准确的方法通常可以应用于蛋白质系统，其中对识别，自我识别或折叠过程的理解是必不可少的。该项目的研究成果在计算机辅助分子设计、结构分析和生物技术等领域具有深远的应用价值。为研究开发的代码和脚本正在以开放源代码的形式传播，并且可以很容易地扩展到广泛使用的分子动力学模拟代码和力场的接口。

Montgomery Pettitt of The University of Texas Medical Branch (UTMB) is supported by an award from the Chemistry of Life Processes Program in the Chemistry Division to create a new set of computational tools for simulating and understanding the fundamental properties of proteins in solution, with application to protein design and the development of new materials and biotechnologies. Some of the most challenging systems in biotechnology applications are those involving liquids, particularly mixtures of biopolymers in water. Professor Pettitt is developing a new computational framework for the molecular design of proteins based on current theories of how liquids behave. Solutions of proteins, which serve as structural components and perform catalytic chemistry, are central to much of biotechnology. This study will enable the rapid screening of proposed protein modifications computationally with greatly improved accuracy compared to existing methods. Codes, scripts, and documentation developed in this project are being made freely-available to the simulation community, and can be interfaced to widely-used molecular dynamics simulation codes. The project is recruiting several students each year for a 10-week internship in computational biology in Professor Pettitt's laboratory, through outreach efforts at local HBCU and Hispanic-serving institutions and the ACS SEED and R. A. Welch Summer Scholar high school programs. There is close cooperation with the UTMB Sealy Center for Structural Biology, which provides a venue for student research presentations at research symposia and conferences, and opportunities for research collaborations using the new methodology and to provide feedback on future enhancements.The mechanisms governing recognition between proteins, ligands and the transition of proteins from their unfolded state to their native state remain as fundamental chemical questions. Those equilibria are governed by free energy differences in solution. The calculation of solvation free energies by explicit computer simulations can be accurate, but is computationally expensive and essentially prohibitive for entire proteins, even on large supercomputing systems. This project is developing precomputed solvent distributions ("proximal (radial) distribution functions", or pDFs) extracted from benchmark all-atom molecular dynamics simulations to produce a near neighbor approximation to the solvation density profile about arbitrary protein solutes. The initial libraries of pDFs are being developed for the widely-used CHARMM force field. The principal goals of the research are to (1) generate pDF libraries for all naturally-occurring amino acids and additional important functional groups, (2) develop efficient methods based on the pDFs to compute total and relative solvation free energies, and (3) extend the method to multicomponent solutions. The third goal includes experimental collaboration to test the methodology against solution thermodynamic measurements. Given the precomputed nature of the pDFs, the evaluation of thermodynamic averages as integrals over the precomputed distribution is computationally very efficient. These fast and accurate methods can be generally applied to protein systems where the understanding of the recognition, self-recognition, or folding process is essential. This results of this project have far-reaching applications to the fields of computer-aided molecular design, structural analysis, and biotechnology. The code and scripts developed for the research are being disseminated as open source and can be readily extended to interface to widely-used molecular dynamics simulation codes and force fields.