Discovering the molecular genetic principles of cell type organization through neurobiology-guided computational analysis of single cell multi-omics data sets

摘要 理解细胞类型多样性和组织的生物学原理对于破译 大脑功能的神经回路近年来，单细胞转录组和 表观基因组数据集提供了前所未有的机会，探索细胞的分子遗传基础， 类型身份、多样性和组织。然而，多组学数据集的分析在很大程度上是由 通过统计方法，通常不涉及神经生物学和发育的深入知识， 生物学因此，大多数统计方法不区分技术噪音和方法论偏差， 生物学相关的信号和关系，并且在实现生物学发现方面具有有限的力量， 洞察力.基于固有生理和发育的神经生物学引导的特征选择 过程是必不可少的，以超越简单的分子类型的统计聚类， 神经元类型的多模态定义，并揭示它们作为分类学的内在关系。我们有 发现突触输入/输出（I/O）通信的转录结构可能是 皮质GABA能神经元身份的本质。我们假设突触的转录结构 通信是脑神经元类型的一般定义特征。我们将通过以下方式检验这一假设： 执行一系列的监督学习和特征选择分析， 来自BRIAN Initiative Cell的跨大脑区域和系统的新兴单sc转录组数据集 人口普查网。我们进一步假设，突触的细胞类型的转录特征， 通讯是由根植于epignomic景观的明确定义的基因调控程序精心安排的。 我们将通过联合分析sc-transcriptome、ATACseq和DNA甲基化组数据集来验证这一假设。 来自BICCN的相同皮质细胞群，以鉴定共表达的基因特征， 细胞身份，重点是突触通信的签名。基于转录组学和表观基因组学 结构的突触通信，我们将进一步发展神经生物学引导的特征选择 算法，以改善和完善目前的统计聚类皮质转录组类型。在 此外，我们将生成用于转录组细胞类型自动分类的网络门户工具。我们的研究 将建立一个神经元细胞类型组织的统一范式，其中epignomic景观配置 核心基因调控程序，以塑造定义主要神经元的突触通信特性 类型这项工作将为理解神经元多样性建立一个分子遗传学框架 并实现跨大脑区域和哺乳动物物种的生物分类。

ABSTRACT Understanding the biological principles of cell type diversity and organization is necessary for deciphering neural circuits underlying brain function. The recent rapid accumulation of single cell transcriptomic and epigenomic data sets provides unprecedented opportunity to explore the molecular genetic basis of cell type identity, diversity, and organization. However, analysis of multi-omics datasets have been largely driven by statistic methods that typically do not engage the deep knowledge of neurobiology and developmental biology. As such, most statistic methods do not distinguish technical noise and methodological biases from biologically relevant signals and relationships, and have limited power in achieving biological discovery and insight. Neurobiology guided feature selection based on inherent physiological and developmental processes is essential to move beyond simple statistical clustering of molecular types towards achieving multi-modal definition of neuron types and revealing their inherent relationships as a taxonomy. We have discovered that transcriptional architectures of synaptic input/output (I/O) communication may underlie the essence of cortical GABAergic neuron identity. We hypothesize that transcriptional architectures of synaptic communication is a general defining feature for brain neuron types. We will test this hypothesis by performing a series of supervised learning and feature selection analyses of publically available and emerging single sc-transcriptomic data sets across brain areas and systems from the BRIAN Initiative Cell Census Network (BICCN). We further hypothesize that cell type transcriptional signatures of synaptic communication is orchestrated by well-defined gene regulatory programs rooted in epignomic landscape. We will test this hypothesis by joint analysis of sc-transcriptome, ATACseq, and DNA methylome dataset of the same cortical cell populations from BICCN to identify co-expressed gene signatures that reliably define cell identity, focusing on signatures of synaptic communication. Based on transcriptomic and epigenomic architecture of synaptic communication, we will further develop neurobiology guided feature selection algorithms to improve and refine the current statistical clustering the cortical transcriptomic types. In addition, we will generate web portal tools for automated classification of transcriptomic cell types. Our study will establish a unified paradigm of neuronal cell type organization in which epignomic landscape configures core gene regulatory programs to shape synaptic communication properties that define cardinal neuron types. Together, this work will establish a molecular genetic framework for understanding neuronal diversity and achieving a biological classification across brain areas and mammalian species.