Extension of the Condition-Adaptive Fused Graphical Lasso and Application to Modeling Brain Region Co-Expression Networks

• Activity Number: 55 - Statistical methods for data from single cell technologies
• Type: Contributed
• Date/Time: Sunday, August 8, 2021 : 3:30 PM to 5:20 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Souvik Seal* and Katerina Kechris
• Companies: Department of Biostatistics and Informatics, CSPH, CU Anschutz Medical Campus and University of Colorado Anschutz Medical Campus
• Keywords: Gene co-expression; Gaussian Graphical Model; Fused Graphical Lasso; Condition adaptive Fused Graphical Lasso; ADMM; 1-D variation

Abstract

A gene co-expression network is an undirected graph, where each node corresponds to a gene, and gene pairs are connected with an edge if they have a significant co-expression relationship. Gaussian graphical modeling has been a standard tool for constructing gene co-expression networks. Jointly analyzing gene co-expression profiles under multiple different conditions (e.g., tumor and normal tissue) can significantly increase the power of such an analysis. In addition, examining condition-specific patterns of co-expression can provide insights into the underlying cellular processes activated in a particular condition. Condition-adaptive Fused Graphical Lasso (CFGL) incorporates condition specificity in the estimation of multiple co-expression networks. However, the current implementation of CFGL in R can only accommodate three different conditions and is prohibitively slow for a moderate number of genes. We have developed a C++/Python-based software, Rapid CFGL (RCFGL), that can handle more than three conditions and is computationally more feasible. We have applied RCFGL to examine gene expression data from five different brain regions of a collection of heterogeneous stock rats.