Data-driven, biologically constrained computational model of the hippocampal network at full scale

大脑中的信息处理是由不同神经元类型、神经元形态和网络连接拓扑的内在生物物理属性的复杂相互作用组织和促进的。这些特性导致了特定类型的网络振荡和管理神经信息编码和交换的其他动态过程。该项目旨在以前所未有的规模绘制一幅详细的图景，展示正常情况下海马主神经元和中间神经元的固有属性如何定义网络活动，以及在癫痫条件下这些属性的病理变化如何扰乱海马功能。本活动的首要目标是构建1：1比例、真实和详细的大鼠海马齿状回网络的计算模型，并研究生理和病理生理网络动力学。该项目将使用Blue Waters超级计算机来执行这个详细的模型。此外，该项目旨在通过提供必要的软件基础设施来在神经科学界实现更广泛的影响，以将前所未有的千万亿级计算能力提供给神经科学家，无论他们在高性能计算(HPC)方面的专业知识如何。对GPU的支持和在广泛使用的Neuron模拟器中开发的Intel MIC体系结构将使神经科学家能够充分利用超级计算加速器体系结构，从而促进更高的研究生产率，并为更广泛地使用大规模和全面的大脑详细建模铺平道路。此外，作为这项提议背后的研究目标的一部分而开发的神经科学模型代码和模拟支持软件将向公众开放，并将消除建立模拟和管理结果数据的障碍，这些障碍传统上阻碍了高性能计算资源的利用。

Information processing in the brain is organized and facilitated by the complex interactionsof intrinsic biophysical properties of distinct neuronal types, neuronal morphology, and networkconnection topology. These properties give rise to specific types of network oscillationsand other dynamic processes that govern neural information encoding and exchange. This project is designed to create a detailed picture at unprecedented scale of how theintrinsic properties of hippocampal principal neurons and interneurons define the networkactivity under normal conditions, and how pathological changes in those properties under epilepticconditions disrupt hippocampal function. The overarching goal of this activity is to constructa 1:1 scale, realistic and detailed computational model of the CA1-CA3-dentate gyrus networkin the rat hippocampus and study physiological and pathophysiological network dynamics. The project will use the Blue Waters supercomputer to execute this detailed model.Additionally, the project aims to achieve broader impact in the neuroscience community by providing thesoftware infrastructure necessary to put unprecedented petascale computing capability withinreach of neuroscientists regardless of their expertise in high-performance computing (HPC).The support for GPUs and the Intel MIC architecture being developed in the widely-used NEURONsimulator will enable neuroscientists to fully utilize supercomputing accelerator architecturesand thus catalyze greater research productivity and pave the way for wider use of large-scaleand full-scale detailed modeling of the brain.Furthermore, the neuroscience model code and simulation support software being developed as part of theresearch goals behind this proposal will be made publicly available and will remove thebarriers of setting up simulations and managing results data that traditionally have impededthe utilization of high-performance computing resources.