PC1-guided transcriptomic stratification reveals hepatic transcriptional heterogeneity and defines a myeloid-associated 20-gene signature

Abstract

Hepatic lipid-associated inflammation contributes to metabolic liver disease and cardiometabolic complications. Treatment-based transcriptomic comparisons can obscure inter-individual heterogeneity when animals exposed to the same experimental condition show divergent molecular responses. In the public hyperlipidemic liver transcriptomic dataset GSE338111, conventional sex-adjusted comparison of Amlexanox versus DMSO identified only 21 differentially expressed genes at FDR < 0.05 and |log2FC| [≥] 1, and submission of this DEG set to Metascape yielded no GO Biological Process enrichment result. We therefore applied treatment-independent, PC1-guided transcriptomic stratification based on the 500 most variable genes. This analysis resolved three PC1-derived groups and enabled derivation of a myeloid-associated 20-gene signature from the G2-versus-G1 contrast. Independent bulk-transcriptomic cohorts supported responsiveness of the signature to dietary challenge and pharmacologic intervention, while single-cell analysis localized its expression predominantly to hepatic myeloid populations. Human cis-eQTL Mendelian randomization and colocalization further identified TAGLN2 as the signature gene with the strongest genetic support for coronary heart disease. Together, these findings show that PC1-guided stratification can improve resolution of heterogeneous hepatic transcriptional responses and provide a cross-cohort molecular signature for subsequent mechanistic and translational evaluation.