New Concepts and Algorithms for Coarse-Graining in Self-Assembling Systems

加州大学圣巴巴拉分校的M.Scott Shell得到了化学部化学理论、模型和计算方法计划的支持，该计划旨在开发用于系统粗粒化的广泛的多尺度算法。该项目由化学、生物工程、环境和运输系统(CBET)分部的界面过程和热力学计划共同资助。在理论思想和数值技术方面的重大进步是许多大长度和大时间尺度问题的定量建模的关键。这项工作利用了一种名为相对熵的新理论概念，该概念量化了粗粒化过程中丢失的信息，并为多尺度问题提供了一种通用的统计机械方法。这个PI正在使用这个框架创建健壮的算法，提供新的粗粒化范例，包括解决高复杂性多参数模型、多体交互、可转移性和CG架构设计的能力。这些算法正在被实现到一个免费可用的、社区开发的粗粒化包中。谢尔博士还使用这一新一代技术来模拟和了解二苯丙氨酸多肽自组装成中空纳米管的早期结构、热力学和分子机制。该项目正在与针对一系列学生的几项教育活动密切结合。对生物分子、材料和许多其他系统的原子分辨率模拟提供了对它们行为的重要洞察和对它们性质的有用预测。然而，这种方法严重受限于最小和最简单的系统的计算成本。为了将建模扩展到更复杂的情况，长期以来，人们习惯于追求粗粒度模型，这些模型删除了原子细节的一小部分，同时大大减少了计算需求。如果一个人能够通过历史上的反复试验或理性的洞察来确定粗粮的“正确”方式，那么这一策略就可能成功。谢尔博士引入了一种新的方法，通过量化和最小化粗粒化造成的信息损失，以一种物理上有洞察力的方式自动执行这一过程。他的团队正在开发强大的新的通用算法，以产生优化和准确的粗粒度模型，使复杂系统模拟的复杂性达到新的水平。特别是，他正在使用这些方法来理解最近发现的一种设计肽，这种肽可以自组装成空心纳米管，在纳米材料和能源方面的应用越来越多。

M. Scott Shell of the University of California Santa Barbara is supported by an award from the Chemical Theory, Models, and Computational Methods program in the Chemistry division to develop broad-based multiscale algorithms for systematic coarse-graining. The project is cofunded by the Interfacial Processes and Thermodynamics program in the Division of Chemical, Bioengineering, Environmental, and Transport Systems (CBET). Major advances in theoretical ideas and numerical techniques are essential to enabling quantitative modeling of many large length and time scale problems. This work leverages a new theoretical concept called the relative entropy that quantifies the information lost during coarse-graining and that provides a general statistical mechanical approach to multiscale problems. This PI is using this framework to create robust algorithms that offer new coarse-graining paradigms, including the ability to address high-complexity many-parameter models, multi-body interactions, transferability, and design of CG architectures. These algorithms are being implemented into a freely-available, community-developed coarse-graining package. Dr. Shell is also using this new generation of techniques to model and understand the early structures, thermodynamics, and molecular mechanisms governing the self-assembly of the diphenylalanine peptide into hollow nanotubes. The project is being closely integrated with several educational activities targeting a range of students. Atomic-resolution simulations of biological molecules, materials, and many other systems offer major insights into their behavior and useful predictions of their properties. However, such approaches are severely limited by computational expense to the smallest and simplest of systems. To extend modeling to more complicated cases, it has long been customary to pursue coarse-grained models that remove a fraction of atomic detail while greatly reducing the computational demands. This strategy can be successful if one can identify, historically by trial-and-error or rational insight, the "right" way to coarse-grain. Dr. Shell has introduced a new approach that automates this process in a physically insightful way, by quantifying and minimizing the information loss due to coarse-graining. His group is developing powerful new and general algorithms that produce optimized and accurate coarse-grained models, enabling a new level of sophistication in simulating complex systems. In particular, he is using these approaches to understand a recently discovered designer peptide that self-assembles into hollow nanotubes with a fast-growing list of applications in nanomaterials and energy.