Construct Cell Specific Networks from Single Cell Expression

• Activity Number: 179 - Statistical Methods in Single-Cell Transcriptomics
• Type: Contributed
• Date/Time: Tuesday, August 4, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Xuran Wang* and Mingyao Li and Edward B. Lee and David Choi and Kathryn Roeder
• Companies: Carnegie Mellon University and University of Pennsylvania and University of Pennsylvania and Carnegie Mellon University and Carnegie Mellon University
• Keywords: single cell; gene network; non-paramatric; statistical genomics; cell specific network

Abstract

Gene networks support modelling and interpretation of biological processes, and single-cell RNA sequencing (scRNA-seq) provides an exciting opportunity for biological breakthroughs. However, due to the sparsity and heterogeneity of the data, it is challenging to construct gene networks. We propose an alternative method by estimating cell specific networks (CSNs). Building on the ideas of Dai et al. (2019), our method constructs a binary gene-gene adjacency matrix for each cell from scRNA-seq data based on a non-parametric test of independence between genes. Although CSNs are estimated with considerable noise, average CSNs can provide better estimates of network structure than traditional gene correlation measures. We apply our method to scRNA-seq data from human brains and analyze CSNs for microglia cells. Cell clustering algorithms and marker gene analysis identify 4 microglia subtypes. Differences between these subtypes are confirmed by significant differences in their CSNs. Using averaged subtype specific CSNs, gene communities are detected and gene flows between subtypes are investigated. We conclude that CSNs provide an alternative approach to identifying novel cell types.