Dissemination of a tool for data-driven multiscale modeling of brain circuits

摘要 标题：大脑回路的数据驱动多尺度建模工具的传播。 少年派：S·杜拉-贝纳尔 我们正在开发一种新的软件工具，称为NetPyNE，它使用户能够合并复杂的实验 将来自不同尺度的数据转换成统一的fi计算模型。然后，用户可以对此进行模拟和分析 模型在一个独特的框架中更好地了解大脑结构、动力学和功能，该框架结合了： 1.使用fl灵活的、基于规则的高级标准化规范fi阳离子进行程序化或图形用户界面驱动的模型构建； 2.将模型参数与底层技术实施分开，防止编码错误和 模型更易于阅读、修改、共享和重复使用；3.支持从分子到细胞再到网络的多尺度； 4.支持复杂的亚细胞机制、树突连接和刺激模式；5.Effi并行 在独立计算机和超级计算机上的模拟；6.自动数据分析和可视化(例如， 连通性、神经活动、信息理论分析)；7.向/从多个标准化的 格式；8.使用网格搜索和进化算法自动调整参数(分子到网络级别)。 几种同行评议的出版物和以下几种方式证明了NetPyNE使fi研究社区受益的潜力 用户和倡导者稳步增长。我们实验室和合作者实验室的50多名研究人员和学生 使用原型工具进行教育或研究各种大脑区域和现象。有一个 活跃的在线社区，他们协力为项目做出贡献，通过 GitHub平台、一个邮件列表和两个问答论坛。计算神经科学组织包括一个 在他们的2019年冬季通讯中关于NetPyNE的2页专题文章。NetPyNE也在与其他 神经科学界的资源：人类新皮质神经解算器、开放源码大脑、神经科学 网关，以及NeuroML和Sonata国际标准化网络格式。 我们的建议旨在将NetPyNE转变为一个可靠且经过良好测试的工具，具有功能齐全的图形用户界面，并广泛 在Sciencefic社区中传播该工具。该工具的快速发展意味着许多功能已经 在资源和时间有限的情况下，快速增加。我们现在将确保对所有这些功能进行适当的评估 可靠性、健壮性和可伸缩性，有很好的文档记录，并整合到图形用户界面中。图形用户界面也将得到扩展 提供基于Web的在线访问并支持较大模型的可视化。我们还将开发互动 在线教程，以清楚地解释和演示我们的包中包含的丰富和多样化的功能。 通过每年为期数天的课程和神经科学会议的教程/研讨会，我们将参与和培训 学生、实验和计算神经学家以及临床医生使用NetPyNE进行多尺度神经研究 模特儿。多尺度建模是对实验的补充，它结合了前面的内容并使其具有可解释性 不可通约的数据集。使用NetPyNE开发的模拟和分析提供了一种更好地理解 跨脑尺度的相互作用，包括分子浓度、细胞生物物理学、电生理学、神经 动力学、群体振荡、脑电/脑磁图信号和信息论措施。

Summary Title: Dissemination of a tool for data-driven multiscale modeling of brain circuits. PI: S Dura-Bernal We are developing a novel software tool, called NetPyNE, that enables users to consolidate complex experimental data from different scales into a unified computational model. Users are then be able to simulate and analyze this model to better understand brain structure, dynamics and function in a unique framework that combines: 1. programmatic or GUI-driven model building using flexible, rule-based, high-level standardized specifications; 2. separation of model parameters from underlying technical implementations, preventing coding errors and making models easier to read, modify, share and reuse; 3. support for multiple scales from molecule to cell to network; 4. support for complex subcellular mechanisms, dendritic connectivity and stimulation patterns; 5. efficient parallel simulation both on stand-alone computers and supercomputers; 6. automated data analysis and visualization (e.g., connectivity, neural activity, information theoretic analysis); 7. importing and exporting to/from multiple standardized formats; 8. automated parameter tuning (molecule to network level) using grid search and evolutionary algorithms. NetPyNE's potential to benefit the research community is evidenced by several peer-reviewed publications and by the steady growth of users and advocates. Over 50 researchers and students in our lab and collaborators' labs have used a prototype of the tool for education or to investigate a variety of brain regions and phenomena. There is an active online community who collaboratively contribute to the project, post questions and request features via the GitHub platform, a mailing list and two Q&A forums. The Organization for Computational Neuroscience included a 2-page feature article on NetPyNE in their 2019 Winter Newsletter. NetPyNE is also being integrated with other resources in the neuroscience community: Human Neocortical Neurosolver, Open Source Brain, Neuroscience Gateway, and the NeuroML and SONATA international standardized network formats. Our proposal is aimed at transforming NetPyNE into a solid and well-tested tool with a fully-featured GUI, and widely disseminating the tool among the scientific community. The rapid growth of the tool means many features have been added at a fast pace, with limited resources and time. We will now ensure all these features are properly evaluated for reliability, robustness and scalability, well documented and incorporated into the GUI. The GUI will also be extended to provide online web-based access and support visualization of larger models. We will also develop interactive online tutorials to clearly explain and demonstrate the ample and diverse functionality included in our package. Through a yearly multi-day course and tutorials/workshops at neuroscience conferences we will engage and train students, experimental and computational neuroscientists, and clinicians in using NetPyNE for multiscale neural modeling. Multiscale modeling complements experimentation by combining and making interpretable previously incommensurable datasets. Simulations and analyses developed with NetPyNE provide a way to better understand interactions across the brain scales, including molecular concentrations, cell biophysics, electrophysiology, neural dynamics, population oscillations, EEG/MEG signals, and information theoretic measures.