Statistical Methods for Single Cell Regulomics

• Activity Number: 480 - Novel Statistical Methods for Bioinformatics and Computational Biology
• Type: Invited
• Date/Time: Wednesday, July 31, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Sunduz Keles* and Daniel Conn
• Companies: UW Madison and University of Wisconsin
• Keywords: scRNA-seq; networks; transcription factors; random forests; zero inflation; count data

Abstract

Single-cell RNA-sequencing (scRNA-seq) emerged as a revolutionary tool that enables addressing scientific questions that were elusive just a few years ago. A key area of study with scRNA-seq is coordinated gene expression within regulatory networks. Regulatory networks are critically important for development and function where aberrations have been associated with numerous diseases. While a number of methods have been developed recently for regulatory network reconstruction of single-cell RNA-seq data, these methods heavily rely on methods for bulk RNA-seq. We specifically consider the class of methods that utilize random forests/regression trees as a building block of regulatory networks and benchmark its properties for scRNA-seq data. We propose zero inflated random forests to better characterize the regulatory relationships of transcription factors and their targets. We illustrate the properties of the ZIRF on scRNA-seq study of progenitor blood cells.