Enabling long timestep molecular dynamics on parallel CPU and GPU environments

描述（由申请人提供）：要求支持开发和实施新的计算方法，用于模拟蛋白质的长时间步长动力学，并应用这些方法研究柔性蛋白质中运动和生物功能之间的关系。这将满足对高保真方法的迫切需求，这些方法可以达到微秒到毫秒的模拟生物相关时间尺度，而不是通常可用的纳秒到微秒的模拟。虽然构象变化本身已经被计算研究了多年，但我们提出的工作与其他方法的不同之处在于：（1）高性能地实现了长时间步长动力学和全原子分辨率下增强采样的新方法;（2）应用这些详细的方法来解决生物分子的灵活性和功能问题。由于定量比较实验是至关重要的测试和更大的影响，我们的计算方法，实验合作提出。这些都以支持信的形式记录下来。这个项目将对NIH资助的研究人员产生广泛的影响，因为目标并行软件包已经有了一个庞大的用户群，并且有一个开放的代码源。更长的MD模拟将允许在蛋白质折叠、蛋白质工程、酶设计、柔性靶标药物设计以及蛋白质和核酸复合物之间的相互作用等领域进行以前不可能的研究。 公共卫生相关性：该提案寻求支持的数值方法，允许模拟蛋白质的构象变化在毫秒的时间尺度，这已经允许百倍的速度比传统的分子动力学的蛋白质。随着对方法的改进，GPU和并行CPU的实现以及广泛使用的软件的公开传播，蛋白质折叠，变构和蛋白质工程的更复杂和强大的研究将成为可能。这些发展将由对调节细胞周期的重要蛋白质的突变体的工程改造的联合实验-模拟研究指导，该蛋白质是阿尔茨海默病和癌症的潜在靶点。

DESCRIPTION (provided by applicant): Support is requested for the development and implementation of novel computational methods for simulating the long timestep dynamics of proteins and the application of these methods to study the relation between motions and biological functions in flexible proteins. This will satisfy an urgent need for high-fidelity methos that reach biologically relevant timescales of microseconds to milliseconds of simulation, rather than the nanoseconds to microseconds simulations that are commonly available. While conformational change has itself been studied computationally for many years, our proposed work differs from other approaches in (1) the high-performance implementation of novel methods for long timestep dynamics and enhanced sampling at all-atom resolution and (2) the application of these detailed methods to address questions of flexibility and function in biomolecules of biomedical relevance. Since a quantitative comparison to experiment is critical for both the testing and greater impact of our computational methods, experimental collaborations are proposed. These are documented by letters of support. This project will have a widespread impact on NIH-funded researchers because the target parallel software packages already have a large user base and have an open code source. Longer MD simulations will allow previously impossible studies to be carried out in the fields of protein folding, protein engineering, enzyme design, drug design to flexible targets, and interactions among protein and nucleic acid complexes. PUBLIC HEALTH RELEVANCE: This proposal seeks support for numerical methods that allow the simulation of protein conformational changes in the millisecond timescale, which already allow hundred-fold speedups over traditional molecular dynamics for proteins. With proposed improvements to the methods, implementation in GPUs and parallel CPUs and public dissemination in widely used software, more complex and powerful studies of protein folding, allostery, and protein engineering will be possible. These developments will be guided by the combined experimental - simulation study of the engineering of mutants of a protein of importance in regulating the cell cycle that is a potential target for Alzheimer's disease and cancer.