PARALLEL SIMULATION OF LARGE SCALE NEURONAL MODELS

大规模神经元模型的并行模拟

• 批准号: 2379407
• 负责人: NIGEL H GODDARD
• 金额: $ 22.22万
• 依托单位: CARNEGIE-MELLON UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 1997
• 资助国家: 美国
• 起止时间: 1997-09-25 至 2002-04-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/2379407
• 关键词: artificial intelligence; biomedical automation; biotechnology; cerebellar cortex; computational neuroscience; computer network; computer program /software; computer simulation; computer system design /evaluation; hippocampus; mathematical model; neural information processing; neurotransmitters; parallel processing; supercomputer; thalamocortical tract; vocabulary development for information system

DESCRIPTION (Taken from application abstract): Over the last decade computational modeling has become central to neurobiology. While much of this work has focused on cellular and sub-cellular processes, the last few years have seen increasing interest in systems level models and in integrative accounts that span data from the subcellular to behavioral levels. Our proposal, in summary, is to extend existing work in parallel discrete event simulation (PDES) and integrate it with existing work on compartmental modeling environments, to produce a software environment which has comprehensive support for modeling large scale, highly structured networks of biophysically realistic cells; and which can efficiently exploit the full range of parallel platforms, including the largest parallel supercomputers, for simulation of these network models, which integrate information about the nervous system from sub-cellular to the whole-brain level. Because of the scale of the models needed at this level of integration, advanced parallel computing is required. The critical technical insight upon which this work rests is that neuronal modeling at the systems level can often be reduced to a form of discrete event simulation in which single cells are node functions and voltage spikes are events. Three neuroscience modeling projects, will mold, test, and utilize these new capabilities in investigations of system-level models of the nervous system which integrate behavioral, anatomical and physiological data on a scale that exceeds current simulation capabilities. In collaboration with computer scientists at Pittsburgh Supercomputing Center and UCLA, neuroscientists at University of Virginia, the Born-Bunge Foundation, Antwerp, and the Salk Institute, and developers of the NEURON and GENESIS packages, these tools will be developed and made available to the neuroscience community. The software development aims include 1) investigation of a portable, PDES system capable of running efficiently on diverse parallel platforms, 2) development of interfaces to the PDES for NEURON and GENESIS allowing models developed in those packages to be scaled up, 3) investigation of a network specification language for neuronal models, and associated a visualization interface, to facilitate investigation of systems-level models, 4) sufficiently robust and well-documented software for download and installation at other sites. The three neuroscience projects will guide development of the software tools and use the tools for investigation of large-scale models of cerebellum, hippocampus and thalamocortical circuits.

描述（摘自申请摘要）：过去十年 计算建模已成为神经生物学的核心。 虽然很多 这项工作的重点是细胞和亚细胞过程，最后几个 多年来人们对系统级模型和 涵盖从亚细胞到行为数据的综合账户 水平。 总之，我们的建议是并行扩展现有工作 离散事件模拟（PDES）并将其与现有工作集成 分区建模环境，以产生一个软件环境 全面支持大规模、高度结构化的建模 生物物理真实细胞网络；并且可以有效地利用 全系列并行平台，包括最大的并行平台 超级计算机，用于模拟这些网络模型，它集成了 有关神经系统从亚细胞到全脑的信息 等级。 由于这个级别所需模型的规模 集成，需要先进的并行计算。 关键的 这项工作所依据的技术见解是神经元建模 系统级别通常可以简化为离散事件的形式 模拟中，单电池是节点函数，电压尖峰是 事件。 三个神经科学建模项目将塑造、测试和利用这些新的 神经系统系统级模型研究的能力 大规模整合行为、解剖和生理数据 这超出了当前的模拟能力。 与合作 匹兹堡超级计算中心和加州大学洛杉矶分校的计算机科学家， 弗吉尼亚大学、伯恩-邦吉基金会的神经科学家， 安特卫普、索尔克研究所以及 NEURON 和 GENESIS 的开发商 包，这些工具将被开发并提供给 神经科学界。 软件开发目标包括1） 研究能够高效运行的便携式 PDES 系统 不同的并行平台，2）开发 PDES 接口 NEURON 和 GENESIS 允许扩展在这些软件包中开发的模型 上，3）神经元网络规范语言的研究 模型，并关联可视化界面，以方便 系统级模型的调查，4）足够稳健和 文档齐全的软件，可在其他站点下载和安装。 这 三个神经科学项目将指导软件工具的开发 使用工具研究大型小脑模型， 海马体和丘脑皮质回路。