A1221
Title: A Bayesian hierarchical model for estimating cell-type-specific multi-omic networks from bulk-omics data
Authors: Enakshi Saha - University of South Carolina (United States) [presenting]
Abstract: Gene regulatory networks provide crucial insights into how genes and regulatory molecules interact to drive complex traits and diseases. These regulatory interactions exhibit significant variation across cell types. Estimating cell-type-specific networks is essential for understanding biological mechanisms of traits and diseases. While algorithms like inferelator can infer cell-type-specific networks from single-cell omics data, single cell sequencing is expensive and is typically collected from only a few individuals. In contrast, bulk omics data is abundant across large populations, but existing network inference methods for bulk data fail to account for heterogeneity in cell type compositions across samples. A hierarchical Bayesian model is proposed that addresses this limitation by first estimating cell type proportions from bulk RNA-seq data using established deconvolution techniques such as xcell and then using these proportions within a hierarchical framework to decompose co-expression networks estimated from bulk gene expression into cell-type-specific co-expression components. The method is built on a non-negative linear regression with sparsity-inducing horseshoe prior. The method is validated using simulated data and yeast bulk and single cell TF knockout data and demonstrates that the approach successfully recovers cell-type-specific regulatory relationships that are masked in traditional bulk network inference methods.
