ITR: Reduced Basis Methodologies for Computation, Analysis and Visualization of Bio-Molecular Simulations

"PI将研究和开发软件和计算方法，以解决分子动力学模拟中目前最具挑战性的问题之一-大型蛋白质，由溶液中的数万个原子组成，折叠时间至少为一微秒。 这将使用基于MD轨迹奇异值分解（SVD）的缩减基方法来完成。- 开发块SVD更新方案，该块SVD更新方案将使得新的轨迹信息能够被邻接到对先前轨迹的当前截断SVD近似，并且避免必须存储整个轨迹;以及实现缩减基积分器，其将与SVD更新方案协同工作以更快速地计算缩减基仿真。调整为潜在语义索引开发的快速行进算法，开发快速图形查询工具，以定位可能与感兴趣的本地结构匹配的站点-开发I/O支持和可视化功能，以处理将产生的超大型数据操作和表示问题。研究的结果将是一个有效的时间集成方案，可以大大限制存储，但解决多个尺度上的细节。 它将在一个完全溶剂化的蛋白质分子上以微秒的时间尺度进行演示，并且将可视化高阶低频运动。

`The PIs will investigate and develop software and computational methods to address one of the most challenging current problems in molecular dynamics (MD) simulations - large proteins, consisting of tens of thousands of atoms in solution over the physiological range of at least a microsecond of folding time. This will be accomplished using a reduced basis approach based on singular value decomposition (SVD) of the MD trajectory.The work will entail:- developing a block SVD updating scheme that will enable new trajectory information to be adjoined to a current truncated SVD approximation to a prior trajectory and avoid having to store the entire trajectory - developing, analyzing, and implementing a reduced basis integrator that will work in concert with the SVD updating scheme to compute the reduced basis simulation more rapidly- adapting the fast marching algorithms developed for latent semantic indexing to develop the rapid graphical query tools to locate sites that potentially match local structures of interest- developing the I/O support and visualization capabilities to handle the extremely large data manipulation and representation problems that will be generatedThe result of the research will be an efficient time integration scheme that can drastically limit storage and yet resolve detail on multiple scales. It will be demonstrated on a fully solvated protein molecule over a time scale of a microsecond, and the high order, low frequency motions will be visualized.