Leveraging multi-species single cell omic datasets to study the evolution of cell type-specific gene regulatory networks

项目总结 比较功能基因组学为研究物种的分子基础提供了一个强大的框架- 特定的特征。基因调控网络(GRN)，控制精确的上下文特定的表达模式 基因在不同物种间的表型多样化方面发挥着重要作用。这些网络是细胞类型的中心 这是一种特殊的功能，在许多疾病中经常被破坏。然而，基因调控网络的比较 跨物种一直是具有挑战性的，因为缺乏足够数量的跨匹配样本 生物学背景。单细胞基因组技术，如单细胞RNA-seq(scRNA-seq)和atac-seq (scatac-seq)，正在给生物学带来革命性的变化，使研究人员能够描绘几乎所有基因组的活动。 每个单元格中的区域。单细胞基因组研究正在迅速扩展到多个物种，为 前所未有的机会来定义细胞类型及其潜在的基因调控网络并研究它们的 进化论。然而，定义跨物种的细胞类型和细胞特定的GRN的计算方法是 在他们的婴儿期。具体地说，多物种scRNA-seq数据集中的样本通过系统发育相关，然而， 现有的集成方法没有对这些关系进行建模。此外，现有的方法是 仅限于跨物种的一对一关系，这使得研究一些主要来源变得困难 在细胞类型同一性方面的进化创新(例如，复制)。在这个项目中，我们将开发新的 解决两个问题的计算方法：(A)定义细胞类型及其在 来自scRNA-seq和scatac-seq数据集的物种，(B)特定细胞类型的推断和比较分析 来自单细胞RNA-seq和atac-seq数据的跨物种GRN。我们的工具将基于机器学习 方法，即概率图形模型，多任务和多视图学习，以及矩阵因式分解， 提供有原则的框架，以整合跨物种的信息。我们将首先在人类身上测试这些工具 来自我们的合作者和已发表的研究的小鼠scRNA-seq/ATAC-seq数据集。我们将演示 我们的工具在一个新的多物种肾脏scRNA-seq/scATAC-seq数据集上的全部潜力，我们将收集到 研究正常的肾功能以及控制一个肾脏如何恢复的代偿性肾发育。 在手术切除了另一个肾脏之后。我们将确定保守和分散的监管网络，这些网络将 用于用CRISPR和siRNA验证研究的序列和蛋白质调节因子的优先顺序。我们的分析 将揭示GRN如何跨物种进化以及它们如何建立不同的细胞类型的关键见解。我们的 方法和新的数据集将为控制肾脏的分子程序提供关键的见解 可能对肾脏疾病患者产生重大临床影响的结构和功能。我们的方法 将组成一套广泛适用的工具，可以深入了解基因调控和细胞的原理 将适用于来自不同多细胞系统的单细胞数据集的FATE规范。

PROJECT SUMMARY Comparative functional genomics offers a powerful framework to study the molecular underpinnings of species- specific traits. Gene regulatory networks (GRNs) which control precise context-specific expression patterns of genes play a significant role in diversifying phenotypes across species. These networks are central to cell type specific function and are often disrupted in many diseases. However, comparison of gene regulatory networks across species has been challenging because of the lack of sufficient number of samples across matched biological contexts. Single cell omic technologies, such as single cell RNA-seq (scRNA-seq) and ATAC-seq (scATAC-seq), are revolutionizing biology enabling researchers to profile the activity of nearly all genomic regions in each individual cell. Single cell omic studies are quickly expanding to multiple species providing unprecedented opportunities to define cell types and their underlying gene regulatory networks and study their evolution. However, computational methods for defining cell-types and cell-specific GRNs across species are in their infancy. In particular, samples in a multi-species scRNA-seq dataset are related by a phylogeny, however, existing integration approaches do not model these relationships. Furthermore, existing approaches are restricted to one-to-one relationships across species, which makes it difficult to study some of the major sources of evolutionary innovation (e.g., duplications) in cell type identity. In this project, we will develop novel computational methods to tackle two problems: (a) defining cell types and their lineage relationships across species from scRNA-seq and scATAC-seq datasets, (b) inference and comparative analysis of cell type-specific GRNs across species from single cell RNA-seq and ATAC-seq data. Our tools will be based on machine learning methods, namely, probabilistic graphical models, multi-task and multi-view learning, and matrix factorization, that offer principled frameworks to integrate information across species. We will first test these tools in human and mouse scRNA-seq/ATAC-seq datasets from our collaborators and published studies. We will demonstrate the full potential of our tools on a novel multi-species kidney scRNA-seq/scATAC-seq dataset that we will collect to study normal kidney function as well as compensatory renal growth, which controls how one kidney recovers after surgical removal of another kidney. We will identify conserved and diverged regulatory networks that will be used to prioritize sequence and protein regulators for validation studies with CRISPR and siRNA. Our analysis will reveal key insights into how GRNs evolve across species and how they establish different cell types. Our approaches and novel datasets will provide critical insight into the molecular programs governing kidney structure and function that could have a significant clinical impact for patients with kidney disease. Our methods will constitute a suite of broadly applicable tools that can shed insight into principles of gene regulation and cell fate specification that will be applicable to single cell datasets from diverse multi-cellular systems.