Investigating the microcircuit determinants of neural population activity through comparative analysis of latent dynamics across cortical areas in the mouse

项目摘要 神经科学的一个关键目标是确定微电路结构如何预测电路功能。一个耐人寻味的想法， 在一些理论模型的支持下，微电路组成的变化支持功能 专业化。这一理论得到了观察到的电路中梯度之间的相关性的支持 特性(受体表达密度；抑制性细胞类型)和平均固有特性的测量 记录的大脑皮层区域活动的时间尺度。然而，其他理论表明，观察到的层次结构 在没有微电路变化的情况下，时间尺度可能会出现，无论是在经过严格调整的随机网络中，还是在 前馈网络，每一层对相关的输入集合求和。此外，单个单元格 嵌入复杂、互连的网络中，在多个时间尺度和广泛的 即使在单个区域内，也可以看到单个单元格时间尺度的分布。之前的经验研究依赖于 关于使用单细胞尖峰计数相关性和平均读数来量化活动的时间尺度 皮层脑电(ECoG)等活动，因为直到最近，大规模并行记录 具有细胞类型信息的单个神经元群体以及分析集体的统计方法 动力学并不是广泛可用的。 我们开发了一个基于动态潜变量模型的分析框架来量化时间尺度和 从当地人口的尖峰活动和当地的人口活动测量得出的皮质动力学状态 (局部场势，LFP)。我们建议将这些方法应用到一组公开可访问的录音中 在小鼠的大脑皮质活动中确定(1)大脑皮质的特定特征的变异程度 微电路解释了皮质动力学的变化，(2)特定细胞类型在决定集合体中的作用 动力学，以及(3)跨记录模式的活动模型之间的联系 这项研究基本上是关于定量绘制正常的功能变异图并确定其程度 根据本地电路属性对该变化进行预测。准确地对这种关系进行建模将具有 对我们预测神经动力学如何由微电路结构和 可以扩展到理解由与心理有关的回路变化引起的活动动力学的中断 健康障碍。与微电路结构无关，这项研究的成功将产生 可以准确和定量地绘制大脑皮质动力学变异性的框架 随着时间的推移，在个人或同一个人中，为学习和研究提供了一个宝贵的工具 发展。

Project Summary A key goal in neuroscience is determining how microcircuit structure predicts circuit function. An intriguing idea, supported by some theoretical models, is that variation in microcircuit composition supports functional specialization. This theory has received support from the observation of a correlation between gradients in circuit properties (receptor expression densities; inhibitory cell types) and in measurements of average intrinsic timescales of recorded activity across cortical areas. However, other theories show how observed hierarchies of timescales can emerge in the absence of microcircuit variation, either in critically tuned random networks or in feed-forward networks with each layer summing over correlated sets of inputs. Moreover, individual cells are embedded in complex, interconnected networks, generating correlations on multiple timescales and broad distributions of single-cell timescales even within a single region. Previous empirical investigations have relied on quantifying timescales of activity using single-cell spike count correlations and by averaged readouts of activity such as the electrocorticogram (ECoG), because until recently, massively parallel recordings of populations of single neurons with cell-type information as well as the statistical methods to analyze collective dynamics were not widely available. We have developed an analytic framework based on dynamic latent variable models to quantify timescales and states of cortical dynamics from spiking activity in local populations and from local measures of population activity (the local field potential, LFP). We propose to apply these methods to a set of publicly accessible recordings of cortical activity in the mouse to determine (1) the extent to which variation in specific features of the cortical microcircuitry explains variation in cortical dynamics, (2) the role of specific cell types in determining collective dynamics, and (3) the connection between activity models across recording modalities This study is fundamentally about quantitatively mapping normal variation in function and determining the extent to which that variation is predicted from the local circuit properties. Modeling this relationship accurately will have a large impact on our ability to predict how neural dynamics arise from changes in microcircuit structure and could be extended to understand disruption of activity dynamics arising from circuit changes linked to mental health disorders. Independent of the relationship to microcircuit structure, success of this study will generate a framework in which variability of cortical dynamics can be accurately and quantitatively mapped across individuals or in the same individual over time, providing an invaluable tool for the studies of learning and development.