How Gene Networks Reveal New Genes Linked to Coronary Artery Disease?

Genetic research is reshaping how we understand complex diseases like coronary artery disease (CAD). Traditional transcriptome‑wide association studies (TWAS) link genetic variants to traits by looking at cis‑eQTLs, i.e., nearby effects. Yet DNA works like a city with long roads: many variants act at a distance, influencing other genes through regulatory networks. The new GRN‑TWAS framework rebuilds these gene regulatory networks (GRNs) and feeds them into gene‑expression prediction models, capturing trans‑eQTL effects that travel through the network. Applying the method to 600 CAD patients from the STARNET study and GWAS data, researchers improved expression prediction and uncovered 5,779 significant genes, more than half never linked to CAD before. A ranking engine highlighted 882 highly relevant genes, 237 of which are regulated solely by trans effects. Key‑driver analysis identified 18 potential trans mediators with high network centrality, offering mechanistic hypotheses for association signals. In short, embedding network topology into TWAS provides a full‑picture view: from genotype, through gene interactions, to disease phenotype. This opens new avenues for therapeutic target discovery and a deeper grasp of CAD’s genetic complexity.

The GRN‑TWAS model hinges on two core steps: (1) reconstruction of the gene regulatory network (GRN) and (2) integration of network topology into gene‑expression imputation models. For GRN reconstruction, researchers combined multi‑tissue expression data with inference methods based on partial correlations and LASSO regression, producing directed graphs that depict activation or repression relationships among genes. The topology is then encoded into weighted adjacency matrices, which are introduced as covariates in expression prediction models (e.g., elastic net). This allows the effect of a SNP on a target gene to propagate through arbitrarily long paths, thereby capturing trans‑eQTL effects. Statistically, the framework employs Bayesian regression to estimate weight coefficients, ensuring effective regularization amid many linked variables. Analyses were performed across seven CAD‑relevant tissues (e.g., liver, skeletal muscle, coronary artery), leveraging GWAS summary statistics to compute gene‑phenotype association tests via colocalization methods. Gene prioritization used a ranking algorithm based on evidence scores (p‑value, Z‑score) and network centrality metrics (degree, betweenness). The 18 key drivers emerge as hubs with high connectivity and were in‑silico validated through pathway analyses (KEGG, Reactome) showing enrichment in inflammatory response, lipid metabolism, and vascular signaling processes. This systems‑level approach supports the omnigenic hypothesis, whereby most genes in the genome contribute, directly or indirectly, to phenotypic variation in a complex disease like CAD.