Defining gene regulatory networks controlling cell fate

点击翻译按钮获取中文摘要

PROJECT SUMMARY Cell type-specific transcriptional networks are gene regulatory networks that dynamically reconfigure to drive precise spatio-temporal expression patterns of genes. These networks are central to cell type specificity and are often disrupted in many diseases. The structure of these networks is defined by a trans component that specifies which regulatory proteins control a gene’s expression and a cis component that species the regulatory regions that can regulate a gene’s expression both proximally and distally. Identifying these regulatory networks has been a significant challenge for mammalian cell types because of the number of potential regulators of a gene and the large number of assays needed to define these networks accurately. Advances in single cell omics technologies, such as single cell RNA-seq (scRNA-seq) and single cell ATAC-seq (scATAC-seq), offer new opportunities to define cell type-specific regulatory networks because of their ability to comprehensively profile the transcriptome and accessibility for thousands of individual cells. However, computational methods for integrating these data to define both cell lineage structure and cell-type specific regulatory networks are limited. Most methods have used only one type of assay focusing either on the cis or trans components and have not modeled temporal or hierarchical relatedness of multi-sample datasets. Finally, performance of computational network inference methods has remained low when compared to experimentally detected networks. To address these challenges, we will develop novel computational methods and powerful resources for mapping gene regulatory network dynamics driving cell type specificity. Our aims are to (a) develop a computational toolkit to integrate scRNA-seq and scATAC-seq datasets to infer both cell type lineage (Aim 1) and cell type-specific transcriptional regulatory networks from scRNA-seq and ATAC-seq data (Aim 2), (b) identify the rewired network components during a dynamic progress such as cellular reprogramming (Aim 2), and (c) develop an active learning based approach to infer causal regulatory networks and apply this framework to refine the regulatory networks for cellular reprogramming (Aim 3). We will apply our tools to public and newly collected datasets as part of this project. Our analysis will reveal cis and trans regulatory network components associated with cell fate specification during a dynamic process such as reprogramming or development. Our active learning approach will use Perturb-Seq to perform regulator perturbations to both validate the predicted networks as well as to establish improved gold standards for a system with high significance for translational and basic research. The tools and datasets generated by this project will be publicly available and will serve as a powerful resource to understand regulatory network dynamics in cell fate specification. Our tools should be broadly applicable to define regulatory network dynamics for diverse biological processes.