Collaborative Research: On-the-fly free energy parameterization in molecular simulations

技术总结在这个项目中，由凝聚态物质和材料理论计划资助，提出了一种新的方法，用于从分子动力学（MD）模拟中提取自由能。 该方法基于温度加速，其中在运行的MD模拟中对一些期望的集合变量中的自由能梯度进行采样，同时将这些集合变量驱动越过自由能屏障，否则这些自由能屏障将使传统MD在这种采样时无望。 这些梯度被用来作为输入到一个优化，最大限度地减少与梯度的分析自由能通过演变的自由能？s参数。 因此，该方法将MD系统、人们想知道自由能的任何集体变量的定义以及该自由能的函数形式作为输入，然后该方法产生最佳参数集。 该方法是制定在一个一般的方式，因此，预计将广泛适用于几乎任何所需的集体变量。 该项目将致力于开发这种方法，特别是用于涉及获得有效的粗粒度的多尺度模拟和大生物分子的构象统计研究的潜力的应用。该项目还将为代表性不足的学生提供最先进的计算和数值方法的培训和指导。非技术性总结无论是在设计新药或新材料，还是仅仅试图了解自然界的工作原理，计算机模拟对于理解物质在分子水平上的行为都很重要。 将这种分子水平的理解带入日常感知领域需要统计方法，这些方法并不总是简单明了的，而且往往相当昂贵。 该项目提出了一种新的统计方法，用于从分子水平的模拟中提取信息。 该方法可能比现有方法更有效。 本项目中进行的研究将开发和应用这种方法来测试与生物和材料应用相关的系统。 该项目可以显着降低从化学，物理和生物学交叉点的分子水平模拟中获得高质量信息的成本。该项目还将为代表性不足的学生提供最先进的计算和数值方法方面的培训和指导。这项工作得到了凝聚态物质和材料理论计划的支持。

TECHNICAL SUMMARYIn this project, funded by the Condensed Matter and Materials Theory Program, a new method is proposed for extracting free energies from molecular dynamics (MD) simulations. The method is based on temperature-acceleration, wherein free energy gradients in some desired set of collective variables are sampled in a running MD simulation while simultaneously driving those collective variables over free energy barriers that would otherwise render traditional MD hopeless at such sampling. These gradients are used as inputs into an optimization that minimizes the error associated with gradients of an analytical free energy by evolving that free energy?s parameters. The method therefore takes as input an MD system, the definitions of whatever collective variables in which one would like to know the free energy, and a functional form for that free energy, and then the method produces the optimal set of parameters. The method is formulated in a general way and is therefore expected to be applicable broadly to almost any desired collective variables. This project will be devoted to developing this method particularly for applications involving deriving effective coarse-grained potentials for multiscale simulations and to the study of conformational statistics of large biomolecules. The project will also provide training and mentoring of underrepresented students in state of the art computational and numerical methodsNON-TECHNICAL SUMMARYWhether in designing new drugs or new materials, or simply trying to understand how Nature works, computer simulations are important in understanding how matter behaves at the molecular level. Bringing such molecular-level understanding into the realm of everyday perception requires statistical approaches that are not always straightforward and often quite expensive. This project proposes a new statistical method for extracting information from molecular-level simulations. The method is potentially much more efficient than existing approaches. The research undertaken in this project will develop and apply this method to test systems relevant for biological and materials applications. The project could significantly reduce the cost associated with obtaining high-quality information from simulations at the molecular level that at the intersections of chemistry, physics, and biology. The project will also provide training and mentoring of underrepresented students in state of the art computational and numerical methods. This work is supported by the Condensed Matter and Materials Theory Program.