Modeling the structure-function relation in a reconstructed cortical tissue

摘要 神经元之间的连通性与这些神经元在体内表现出的活动模式有什么关系？这 大脑回路中的结构-功能关系问题具有重要意义。在一个电话中回答它 量化的方式将对我们关于大脑如何工作的理论和 应用范围从更好的疾病治疗到人工智能的新工具。然而，我们目前 对结构-功能关系的理解相对较差，很大程度上是因为 神经元的连接在很大程度上仍然是未知的。反过来，这严重限制了将建模连接到 理论上的努力。这一问题在研究高度异质性的 大脑皮层回路，参与感知、认知和学习等重要功能。 幸运的是，我们在艾伦脑科学研究所的合作者最近的实验工作现在是 从而产生了变革性的新数据集，该数据集表征了 使用多片突触生理学的细胞类型水平和使用电子的单个神经元水平 显微镜(EM)。在神经科学史上，我们将第一次拥有单个神经元的连接体 再加上V1~1mm3活动的密集记录，以及突触的系统特征 属性。 我们将利用这些独特的数据集来构建和与社区共享V1和使用的新模型 研究大脑皮质连通性与体内活动和计算之间的关系。我们会 分析神经元编码的多种特征如何依赖于单个细胞的属性和更高的阶数 连接性主题，存在于EM连接体中，但不存在于基于统计的连接性中 从细胞类型水平的稀疏测量或从现有文献中推断。我们还将评估 V1的新模型与当前结构-功能关系理论的预测的一致性。 这些模型和模拟将作为科学家将提供的资源与社区免费共享 用于指导未来的实验设计，改进模型中的生物现实主义，并帮助生成和 测试理论。通过为新的理论、模型和技术提供丰富的生物现实框架 实验研究，这一资源将推动关于结构和 大脑皮层回路的功能。

Abstract How is connectivity between neurons related to patterns of activity exhibited by these neurons in vivo? This question of structure-function relations in brain circuits is of fundamental importance. Answering it in a quantitative manner would have far-reaching consequences both for our theories of how brain works and for applications ranging from better disease treatments to new tools for artificial intelligence. However, our current understanding of structure-function relations is relatively poor, in large part because the fine structure of neuronal connectivity has remained largely unknown. In turn, this severely limits connecting modeling to theoretical efforts. This problem is particularly challenging in the case of studying the highly heterogeneous cortical circuits, which are involved in important functions like perception, cognition, and learning. Fortunately, recent experimental work by our collaborators at the Allen Institute for Brain Science is now resulting in transformational new datasets that characterize connectivity in the mouse cortical area V1 at the level of Cell Types using multi-patch synaptic physiology and at the level of individual neurons using electron microscopy (EM). For the first time in history of neuroscience, we will have connectome of individual neurons coupled with dense recordings of activity in ~1 mm3 of V1, plus systematic characterization of synaptic properties. We will leverage these unique datasets to build and share with the community new models of V1 and use them to study the relationships between cortical connectivity and in vivo activity and computations. We will analyze how multiple features of neuronal code depend on individual cell properties and on higher-order connectivity motifs, which are present in the EM connectome, but not in the statistics-based connectivity inferred from sparse measurements at the Cell Types level or from existing literature. We also will evaluate the consistency of the new models of V1 with predictions made by current theories of structure-function relations. These models and simulations will be freely shared with the community as a resource that scientists will use to guide future experiment designs, improve biological realism in models, and assist in generating and testing theories. By providing a rich and biologically realistic framework for new theoretical, modeling, and experimental studies, this resource will fuel new discoveries regarding relations between the structure and function of cortical circuits.