Development and Application of a Multi-Scale Markov Model for Simulating Protein Folding

分子模拟的基本挑战之一是，化学细节往往是回答感兴趣的问题所必需的，但正是这种细节极大地限制了研究的可行性。多尺度模拟方法提供了解决这一问题的希望，它们结合了多个尺度的结果，只在最需要的地方使用细节。仍然存在几个关键问题，例如如何知道哪些细节最相关？如何有效地、系统地、严格地结合多个尺度的结果？在这里，PI提出了一种新的、高度灵活的多尺度方法的框架，它使用马尔可夫状态模型(MSM)作为多尺度模拟的框架。这种多尺度的MSM(或MSM)可以结合不同模拟方法的结果，期望得到“两全其美”，即更详细的模型的准确性和更简单的模型的计算可处理性。此外，MSMSM提供了一个自然的统计框架，用于将多种模拟方法结合在一起，从许多较短的模拟中预测长时间尺度的行为，并描述和阐明复杂的、不同种类的动力学。为了测试这些方法，PI将把这个模型应用到蛋白质折叠的研究中。这项工作有几个更广泛的影响。首先，如果成功，PI的方法可能会在分子模拟领域产生非常广泛的影响，而不仅仅是蛋白质折叠，因为理解哪些细节相关以及简单与复杂模型的使用基本上存在于分子模拟的所有领域。其次，由于PI在这个项目中使用Folding@Home分布式计算平台，因此有一个重要的教育和推广部分；PI开创了几个推广方面，包括使用YouTube视频、PlayStation3和大规模分布式计算作为推广，他建议继续这一工作。这项工作产生了影响，使许多普通公众在复杂的水平上熟悉生物和化学，因为许多参与Folding@Home的人希望了解他们正在运行的模拟。此外，PI计划扩大这些领域，通过与当地学校的教师直接互动，建立课程和网页，将其放在我们的Folding@Home网站上，从而纳入参与人数不足的群体的手段。该项目由分子和细胞生物科学部的分子生物物理学和化学部的理论、模型和计算方法计划共同支持。

One of the fundamental challenges in molecular simulation is the fact that chemical detail is often necessary for answering questions of interest, yet it is this very detail which greatly limits what is feasible to study. Multi-scale simulation methods give the hope to solve this problem, by combining results from multiple scales, using detail only where it is needed most. Several key issues remain, such as how can one know which details are the most relevant? How can one efficiently, systematically, and rigorously combine results from multiple scales? Here, the PI presents a framework for a novel and highly flexible multi-scale approach, which uses Markov State Models (MSMs) as a framework for multi-scale simulation. Such multi-scale MSMs (or MSMSMs) can combine results from different simulation methodologies, with the expectation of getting the "best of both worlds", i.e. the accuracy of more detailed models and the computational tractability of simpler models. Moreover, MSMSMs present a natural, statistical framework for bringing together multiple simulation methodologies, predicting long timescale behavior from many shorter simulations, and describing and elucidating complex, heterogeneous dynamics. To test these methods, the PI will apply this model to the study of protein folding. There are several broader impacts of this work. First, if successful, The PI's methods could have very broad impact in molecular simulation in fields very far from just protein folding, since the issues of understanding which details are relevant and the use of simple vs complex models are present in essentially all fields of molecular simulation. Second, as the PI uses the Folding@home distributed computing platform for this project, there is a significant education and outreach component; the PI has pioneered several outreach aspects, including the use of YouTube videos, PlayStation 3, and large-scale distributed computing as outreach, and he proposes to continue that work with this project. This work has had the impact in making many people in the general public familiar with biology and chemistry at a sophisticated level, as many individuals participating in Folding@home wish to understand the simulations they are running. Moreover, the PI plans to expand these areas to incorporate means to engage underrepresented groups by direct interaction with teachers from local schools, building curricula and web pages to be put on our Folding@home web site. This project is jointly supported by Molecular Biophysics in the Division of Molecular and Cellular Biosciences and the Theory, Models and Computational Methods Program in the Chemistry Division.