Large neuronal networks: from structure to function, and back

我们的研究计划的长期目标是发展新的理论和计算方法，将帮助神经科学家破译神经系统的结构和功能之间的复杂关系在不同的水平：细胞，群体和系统。为此，我们将针对三个特定的项目，然后将每个项目中开发的方法应用于真实的神经元系统。首先，我们建议识别对整个网络的特定功能至关重要的神经元和神经元群落。简而言之，如果一个神经元或一个群体的改变导致网络整体活动的剧烈变化并导致功能丧失，那么它就是关键的。因此，第一个项目与结构对其功能的影响有关。其次，我们的目标是开发从活动数据中准确检测因果关系的新方法。对于神经元网络，因果关系与网络结构密切相关，因为只有当一个神经元与另一个神经元连接时，一个神经元才可能影响另一个神经元。因此，我们希望使用神经网络的函数来预测其结构。三是结构和功能相结合。我们的目标是开发新的方法来预测神经网络的弹性，其中结构和功能通过突触可塑性不断相互影响。我们将确定结构（例如，低互易性）和适应性（例如，快速突触增强）的关键特征，这些特征使整个网络不太容易受到扰动（例如，连通性丧失，突触增强）。为了达到我们的目标，我们需要改进当前用于降低复杂动力系统维数的方法，这些方法仅适用于过于简单的网络结构（例如，随机连接）和具有低生物学基础的神经元动力学（例如，泄漏集成和火灾模型）。这就是为什么我们会从统计物理、非线性动力学和光谱分析中借用一些工具的原因。只有具有明确生物物理解释的神经元模型，如霍奇金-赫胥黎模型，将被考虑。在5年期间结束时，我们将有一个独特的理论和计算框架，将允许分析结构-功能关系，并设计有针对性的干预措施，旨在恢复神经网络的良好功能。

The long-term goal of our research program is to develop novel theoretical and computational methods that will help neuroscientists to decipher the intricate relationship between the structure and the function of nervous systems at different levels: cell, population, and system. To do so, we will target three specific projects and then apply the methods developed in each project to real neuronal systems. First, we propose to identify neurons and communities of neurons that are critical to specific functions of the whole network. Briefly, a neuron or a community is critical if its alteration produces dramatic changes in the global activity of the network and results in a loss of function. This first project is thus related to the impact of the structure on its function. Second, we aim to develop new methods that accurately detect causal interactions from activity data. For neuronal networks, causality is closely related to network structure since a neuron may influence another neuron only if the former is connected to the latter. We thus want to use the function of a neuronal network to predict its structure. Third, we will combine structure and function. Our goal is to develop new methods that predict the resilience of neuronal networks in which both the structure and the function constantly influence each other via synaptic plasticity. We will identify the critical characteristics of both structure (e.g., low reciprocity) and adaptation (e.g., fast synaptic potentiation) that make the whole network less vulnerable to perturbations (e.g., loss of connectivity, synaptic potentiation). To reach our goals, we will need to improve the current methods used for reducing the dimension of complex dynamical systems, which work well only for over-simplistic network structure (e.g., random connectivity) and neuronal dynamics with low biological foundation (e.g., leaky integrate-and-fire model). This is why several tools will be borrowed from statistical physics, nonlinear dynamics, and spectral analysis. Only neuron models with a clear biophysical interpretation, such as the Hodgkin-Huxley model, will be considered. At the end of the 5-year period, we will have a unique theoretical and computational framework that will allow to analyze the structure-function relation and to design targeted interventions aimed at restoring the good function of neuronal networks submitted to perturbations.