AF: Small: CCF: CISE: Advanced Grid-Enabled Algorithms for Discovering Protein Conformations

先进的网格算法来发现蛋白质的构象蛋白质是需要移动才能工作的分子机器。在计算上，人们可以根据根据能量函数加权的物理运动方程来模拟它们的运动。发现蛋白质的形状或构象是一个在非常高维空间中的采样问题，因此是困难的。解决这个问题可以帮助发现与生物相关的反应中间体，设计新的药物，并更好地表征它们的生物学功能。这一建议将开发新的算法和软件来发现蛋白质的构象。我们利用分布式或网格计算来追求对构象空间的探索。我们使用降维方法，将搜索限制在蛋白质最慢和最集体的运动上。这大大加快了采样速度，并可扩展到大型蛋白质。我们将在两个算法中应用我们的技术：副本交换和动态字符串方法。副本交换使用不同温度下的多次模拟，并在这些模拟之间进行周期性的交换以实现退火化。弦方法找到了连接蛋白质两种已知构象的最小自由能路径：例如，从非活性酶到活性酶。我们的技术会自动找到？慢？以及用于搜索构象的集体变量。智力优势：我们的方法有望将搜索速度加快3个数量级或更多，为研究较大蛋白质在长时间内发生的大型构象变化开辟了全新的研究途径。广泛影响：这些算法将作为开放源码软件发布，具有Python接口和GPU加速实现，并将能够使用大型分布式系统。验证将在可获得实验数据的特定蛋白质系统上进行。在与Shodor基金会的合作下，我们将让高中生和教师参与开发网络内容，以支持这项提案的工作。我们将探索在课程中包括这些材料是否会提高人们对计算生物学、生物学或计算的兴趣。

Advanced Grid-Enabled Algorithms to Discover Conformations of ProteinsProteins are molecular machines that need to move in order to function. Computationally one can model their motion according to physical equations of motion that are weighted according to an energy function. Discovery of the shapes or conformations of proteins is a sampling problem in a very high dimensional space, and thus difficult. Solving this problem can assist in the discovery of biologically relevant intermediates of reactions, design of new medicinal drugs, and better characterization of their biological function.This proposal will develop new algorithms and software to discover the conformations of proteins. We exploit distributed or grid computing to pursue exploration of the conformational space. We use a dimensionality reduction approach where the search is confined to the slowest and most collective motions of the protein. This greatly accelerates sampling and is scalable to large proteins. We will apply our technique in two algorithms, replica exchange and the on-the-fly string method. Replica exchange uses multiple simulations at different temperatures and performs periodic exchanges among them to achieve annealing. The string method finds a minimum free energy path that connects two known conformations of a protein: for instance, inactive to active enzyme. Our technique automatically finds ?slow? and collective variables along which to search conformations. Intellectual merit: Our methods hold promise to accelerate the search by 3 or more orders of magnitude, opening entirely new avenues of research related to large conformational changes that happen over long timescales for larger proteins.Broader impacts: These algorithms will be released as open source software with a Python interface and GPU-accelerated implementations, and will be able to use large distributed systems. Validation will be done on a specific protein system for which experimental data is available. In collaboration with the Shodor Foundation, we will involve high school students and teachers in developing web content in support of the work on this proposal. We will explore whether including these materials in courses raises interest in computational biology or biology or computation.