Collaborative Research: The Power of Many: Scalable Compute and Data-Intensive Science on Blue Waters

迫切需要比目前更快、更大规模地解决具有社会和技术重要性的问题。该项目将支持几个新的应用程序，以在科学上需要的规模上利用Blue Waters超级计算机，但在其他方面或其他地方根本不可能。这些领域包括从蛋白质构象到极地科学的各种应用。如果成功，这项研究将在走向极端规模计算和数据密集型科学的能力方面迈出重大一步。作为该项目的一部分提供的计算资源将使设计、开发和测试多种新的算法、中间件和方法成为可能。在第一个轨道中，研究人员提出了一种新的方法，通过使用新的采样方法和工作量管理工具来表征不同形式的NMDAR LBD的构象图景。利用复杂的采样方法，模拟将提供一幅前所未有的原子细节图像，展示蛋白质作为与其相互作用的配体的函数可以采用的不同状态，并将提供将这些状态连接在一起的动力学路径的预测。其结果将成为理解有时令人困惑的实验结果的概念框架，并为进一步实验的合理设计提供跳板。这项研究将开辟一种概念上不同的方法来研究复杂大分子的大构象变化，因为相同的方法和计算基础设施原则上可以应用于大量生物医学相关的系统。该项目的第二条轨道涉及SPIDAL(可伸缩并行可互操作数据密集型库)和MIDAS(数据密集型分析和科学中间件)的开发、扩展和优化，它们将用于增强一系列数据密集型应用程序的可扩展性。

There is a critical need to solve problems of societal and technical importance faster and at larger scales than is currently possible. This project will support several new applications to utilize the Blue Waters Supercomputer at scales that are scientifically needed but simply not possible otherwise or elsewhere. These include applications ranging from protein conformations to polar sciences. If successful, this research will engender a significant step up in capabilities towards extreme scale computing and data-intensive science. The computational resources made available as part of this project will enable the design, development and testing of multiple new algorithms, middleware and methods. In the first track, the researchers propose a new approach to characterize the conformational landscape of the NMDAr LBD in its different forms, by using novel sampling methodologies and workload management tools. Leveraging sophisticated sampling methods, simulations will provide an unprecedented atomically detailed picture of the different states the protein can adopt as a function of the ligands it interacts with, and will also provide predictions of the kinetic pathways that link these states together. The result will be a conceptual framework to understand the sometimes perplexing experimental results and a springboard for the rational design of further experiments. This study will open the way to a conceptually different approach to studying large conformational changes in complex macromolecules, as the same methodology and computational infrastructure can in principle be applied to a large number of biomedically relevant systems. The second track of this project is concerned with development, scaling and optimization of SPIDAL (Scalable Parallel Interoperable Data Intensive Libraries) and MIDAS (Middleware for Data-intensive Analysis and Science) that will be used to enhance the scalability of a range of data-intensive applications.