Multi-omics and machine learning-based exploration of key genes associated with abdominal aortic aneurysm

Abdominal aortic aneurysm (AAA) represents a high-risk arterial pathology that frequently evolves insidiously and remains without robust molecular tools for timely detection. To identify potential biomarkers with both genetically supported relevance and discriminatory value, we developed an integrated multi-omics framework that synthesizes genetic, transcriptomic, and proteomic evidence.

Two microarray datasets (GSE47472, GSE57691) were merged following batch correction and validated using GSE7084. To nominate candidate genes and proteins with genetically supported relevance to AAA, we integrated genome-wide expression and protein quantitative trait loci (eQTL and pQTL) summary statistics within a two-sample Mendelian randomization (MR) analytical design. Differentially expressed genes overlapping MR-supported candidates were further refined using multiple complementary machine learning (ML) approaches. Bayesian colocalization was employed to investigate the extent to which gene expression regulators share genetic architecture with AAA susceptibility. Discriminatory performance was evaluated by constructing receiver operating characteristic curves, and expression changes were validated in a calcium chloride-induced murine AAA model through experimental validation.

A set of 551 genes exhibited significant differential expression in AAA tissues relative to non-aneurysmal controls. MR analyses revealed 267 eQTL-supported genes and 129 pQTL-supported proteins associated with AAA risk, yielding 14 overlapping candidates. ML integration consistently prioritized 4 genes-PLAU, CD58, PCYOX1, and THBS4. Among these, PLAU demonstrated strong colocalization with AAA risk loci (PPH₄ = 0.983) and robust discriminatory accuracy across training and validation cohorts (area under the curve = 0.854 and 0.944). Gene set enrichment analysis and immune-infiltration profiling indicated that elevated PLAU expression is associated with complement activation and enrichment of pro-inflammatory immune subsets, including activated dendritic cells, natural killer T cells, regulatory T cells and T helper 17 cells. In vivo validation demonstrated marked elevation of PLAU expression at both transcript and protein levels in aneurysmal aortas, alongside increased circulating PLAU levels in peripheral blood.

This integrative MR-ML-colocalization strategy provides a comprehensive framework for prioritizing potential biomarker genes in AAA. Convergent multi-omics evidence, together with experimental validation, identifies PLAU as a robust candidate gene strongly associated with genetic susceptibility and immune-inflammatory vascular remodeling.