CDS&E: Fast, Accurate Molecular Solvation Theory for Multiscale Modeling

泰勒Luchko的加州州立大学，北岭，是由一个奖项的支持，从化学理论，模型和计算方法计划在化学部门，以发展大规模分子模拟的方法。 先进网络基础设施部门的计算和数据支持科学与工程（CDS E）计划有助于该奖项。 计算建模经常用于在分子水平上理解蛋白质、DNA和各种小分子之间的相互作用。计算方法的使用使我们对基础生物学的理解和新分子（如抗病毒药物）的设计取得了进展。然而，现实的计算机模拟需要支持这些相互作用的水环境的精确模型。考虑每个水分子位置的模型在物理上是准确的，但随着分子数量的增加，所需的计算时间很快变得令人望而却步。其他方法取代了水的分子细节，而且速度快得多，但以精度为代价。三维参考相互作用位点模型（3D-RISM）是第三种方法，通过计算水分子的密度分布来避免遵循水的原子位置。3D-RISM已经成功地用于研究生物学问题，例如DNA周围的盐和水分布以及小分子与蛋白质的结合。该项目旨在通过进一步发展理论来改进3D-RISM，以更好地捕获分子水的压力和密度分布，并应用先进的数值方法来加快这些计算。 Luchko及其同事的目标是涵盖多个长度尺度的问题，例如细胞内结构的自组装。3D-RISM理论的进展与AmberTools分子建模套件一起分发，允许更广泛的研究社区免费访问这些方法。研究对象为本科生和硕士生。该项目的重点是开发三种独立但互补的方法，以增加大规模模拟的细节和准确性，从而实现目前无法实现的分子模拟。改进3D-RISM的基础理论提供了原子溶剂模型的准确性，而无需显式地模拟它们。这使得晶体结构的细化和DNA周围的溶剂分布被确定使用原子溶剂模型在目前所需的计算成本的一小部分。计算效率提高一到两个数量级，将3D-RISM的现有原子细节首次带到由数百万个原子组成的生物分子中。新的终态自由能方法利用了3D-RISM的最新发展和独特功能，为多尺度系统带来更快，更容易和更准确的自由能计算。结合起来，这些进步提高了所有溶剂性质的准确性，将计算时间减少了一到两个数量级，并提高了大型系统的结合自由能预测。这些改进是重要的，因为它们大大增加了原子分子模型可以解决的生物物理问题的范围和规模。

Tyler Luchko of California State University, Northridge, is supported by an award from the Chemical Theory, Models and Computational methods program in the Chemistry division to develop methods for large-scale molecular simulations. The Computational and Data-Enabled Science and Engineering (CDS&E) Program in the Division of Advanced Cyber Infrastructure contributes to the award. Computational modeling is frequently used to understand interactions between proteins, DNA, and a wide variety of small molecules at the molecular level. The use of computational methods has lead to advances in our understanding of fundamental biology and to the design of new molecules, such as anti-viral medications. However, realistic computer simulations require accurate models of the water environment that supports these interactions. Models that consider the position of every molecule of water are physically accurate but the computation time required quickly becomes prohibitive as the number of molecules grows. Other methods replace the molecular detail of water and are much faster, but at the cost of accuracy. The 3D reference interaction site model (3D-RISM) is a third approach that avoids following the atomic positions of water by calculating the density distribution of water molecules. 3D-RISM has already been successfully used to study biological problems, such as the salt and water distribution around DNA and the binding of small molecules to proteins. This project aims to improve 3D-RISM by further developing the theory to better capture the pressure and density distribution of molecular water and apply advanced numerical methods to make these calculations faster. Luchko and co-workers target problems that cover multiple length scales, such as the self-assembly of structures within the cell. Advances in 3D-RISM theory are distributed with the AmberTools molecular modeling suite, allowing free access to these methods for the broader research community. Undergraduates and Master's level students are involved in this research. The focus of this project is to develop three independent but complementary approaches to increase the detail and accuracy of large-scale simulations to enable molecular simulations that are not currently possible. Improving the underlying theory of 3D-RISM provides the accuracy of atomistic solvent models without explicitly simulating them. This allows crystal structure refinement and solvent distributions around DNA to be determined using atomistic solvent models at a fraction of the computational cost presently required. Increasing computational efficiency one to two orders of magnitude brings the existing atomistic detail of 3D-RISM to biomolecules consisting of millions of atoms for the first time. The new end-state free energy method capitalizes on recent developments for and unique features of 3D-RISM, bringing faster, easier and more accurate free energy calculations to systems of multiple scales. Combined, these advances improve the accuracy of all solvent properties, decrease calculation time by one to two orders of magnitude and improve binding free energy predictions for large systems. These improvements are significant because they drastically increase the scope and scale of biophysical problems that atomistic molecular modeling can address.