Collaborative Research: CIBR: Building Capacity for Data-driven Neuroscience Research

实验神经科学的进步正在产生大量高质量、高分辨率的数据，必须对这些数据进行分析，才能揭示对大脑功能的新见解。处理这种数据雪崩给研究人员带来了一个特殊的挑战，这些研究人员利用大规模高分辨率光学显微镜、功能磁共振成像（fMRI）和高密度脑电活动记录等技术来探测大脑回路和系统的结构和功能。加州大学圣地亚哥分校和耶鲁大学之间的这个合作项目的目的是通过增强神经科学网关（NSG）的能力来促进这类研究。NSG是一种现有的网络基础设施资源，最初是为了促进需要高性能计算的项目而开发的，比如大规模的脑电路计算建模。当前的项目将通过整合高通量计算（HTC）和数据管理方面的创新来增强NSG，这些创新是涉及大量数据的研究所需要的，其实施方式将减少或消除需要处理此类数据的科学家所面临的技术和管理挑战。除了支持数据密集型神经科学研究之外，这些新功能将增加NSG在教育中的效用，它已经广泛应用于本科及以上水平的神经科学和生物学教学。网络研讨会、研讨会和各种会议的培训课程将向学生和研究人员介绍NSG的新能力。该项目将提高NSG作为开放和免费资源的科学和社会价值，通过使所有学术机构的学生和研究人员能够访问计算和数据资源，使科学参与民主化。本项目根据从事数据密集型研究的神经科学家提供的实际和预测用例，在NSG中加入了被认为最适合满足神经科学数据处理大规模计算需求的HTC功能。它结合了商业云计算和开放科学网格（OSG）资源，将它们与NSG的能力相结合，向这些HTC资源提交适当的计算工作负载，同时保持NSG的易用性，允许用户无缝地利用这些计算资源。许多利用HTC计算模式的工具通过NSG提供，允许在NSG现有的基于web和编程的用户环境中处理输入数据和检索输出结果。用户还可以灵活地直接使用商业云计算资源上的神经科学建模和数据处理工具的容器化图像。OSG的数据联合功能的集成允许处理公开可用的大型神经科学数据，这些数据可以以可扩展的方式分发到HTC资源。整合各种数据功能，如能够将大数据直接传输到NSG的存储，在NSG用户之间共享数据，多个NSG用户访问和处理数据，使研究人员能够进行广泛的数据驱动的神经科学研究，无论是处理电生理（脑电图，即EEG，脑磁图，即MEG），成像（fMRI）或行为（反应时间，测试准确性）数据，多模态数据的相关性分析，或者机器/深度学习的应用。在整个项目中，保持与用户社区的密切互动，以便在添加新功能和合并资源时获得反馈。这个项目的网站可以在https://www.nsgportal.org/This上找到。奖项反映了美国国家科学基金会的法定使命，并通过基金会的智力价值和更广泛的影响审查标准进行评估，认为值得支持。

Advances in experimental neuroscience are generating large amounts of high-quality, high-resolution data that must be analyzed in order to reveal new insights into how the brain functions. Dealing with this data avalanche poses a special challenge for research that probes the structure and function of brain circuits and systems with techniques such as large scale high resolution light microscopy, functional magnetic resonance imaging (fMRI), and high density recording of brain electrical activity. The aim of this collaborative project between the University of California San Diego and Yale University is to catalyze such research by enhancing the capabilities of the Neuroscience Gateway (NSG), an existing cyberinfrastructure resource that was originally developed to facilitate projects that need High Performance Computing, such as large scale computational modeling of brain circuits. The current project will enhance NSG by incorporating innovations in high throughput computing (HTC) and data management that are required for research involving large amounts of data, implemented in ways that reduce or eliminate the technical and administrative challenges faced by scientists who need to deal with such data. In addition to enabling data-intensive neuroscience research, these new capabilities will increase NSG's utility in education, where it is already widely used in neuroscience and biology instruction at the undergraduate level and higher. Webinars, workshops, and training classes at various conferences will be presented to students and researchers to learn about NSG's new capabilities. This project will increase NSG's scientific and social value as an open and free resource that democratizes participation in science by enabling access to computing and data resources for students and researchers at all academic institutions. This project adds HTC features to NSG that have been judged most suitable to meet the large scale computing needs for neuroscience data processing, based on actual and projected use cases provided by neuroscientists engaged in data-intensive research. It incorporates commercial cloud computing and Open Science Grid (OSG) resources, integrating them with NSG’s ability to submit appropriate compute workloads to these HTC resources while maintaining the ease of use features of NSG that allow users to seamlessly exploit these compute resources,. Many of the tools that utilize HTC computing mode are made available via NSG to allow processing of input data and retrieval of output results within the existing web based and programmatic user environment of NSG. Flexibility is also provided for users to directly use containerized images of neuroscience modeling and data processing tools on commercial cloud computing resources. Integration of OSG’s data federation capability allows processing of publicly available large neuroscience data which can be distributed in a scalable manner to HTC resources. Incorporation of various data functionalities such as the ability to transfer large data directly to NSG’s storage, share data among NSG users, access and process data by multiple NSG users, enable researchers to perform a wide diversity of data-driven neuroscience research be it processing of electrophysiological (electroencephalography i.e. EEG, magnetoencephalography i.e. MEG), imaging (fMRI) or behavioral (reaction time, test accuracy) data, correlational analysis of multimodal data, or application of machine/deep learning. Throughout the project close interaction with the user community is maintained to gain feedback as new features are added and resources are incorporated. The web site for this project can be found at https://www.nsgportal.org/This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.