Identifying High-Dimensional Genomic Factors Modulating Biological Networks Across Multi-Omics Data

Abstract

Biological traits such as genes, metabolites, and microbial taxa interact within complex networks, yet how genomic factors shape these interactions remains poorly understood. Here, we introduce GFBioNet, a computationally efficient method for identifying factors that modulate direct associations between biological traits within network models. Our two-stage strategy first estimates a baseline network using Gaussian graphical models and then tests whether genomic factors modulate specific network edges (trait-trait relationships), enabling scalable analysis of high-dimensional multi-omics data while explicitly controlling the false discovery rate (FDR). Simulations demonstrate reliable FDR control and high statistical power across a broad range of settings. Applied to multiple datasets, GFBioNet reveals host genetic variants influencing oral microbiome relationships, gut microbial taxa modulating metabolite networks in colorectal cancer, and somatic mutations and copy-number alterations reshaping gene expression networks in lung adenocarcinoma. By expanding network analysis to evaluate modifiers of trait-trait relationships, GFBioNet offers a versatile tool for uncovering the genomic architecture of biological networks across multi-omics studies.