Machine Learning-Based Identification of Exosome-Associated Diagnostic Biomarkers in Acute Myocardial Infarction and...

Acute myocardial infarction (AMI) is a major global health burden, with current diagnostic biomarkers often limited by delayed elevation and low specificity. Exosomes have emerged as promising non-invasive biomarkers for cardiovascular diseases.

Transcriptomic datasets from the Gene Expression Omnibus were integrated with candidate exosome-related gene sets to identify differentially expressed exosome-related genes (DEEGs) in AMI. Functional enrichment analyses were performed to explore biological functions. Machine learning algorithms including LASSO, Random Forest, and SVM-RFE were applied for feature selection. A diagnostic classification model was constructed and its discriminatory performance was evaluated. In silico drug prioritization was performed using the DSigDB database and molecular docking analysis. Plasma-derived exosomes from AMI patients and healthy controls were isolated for experimental validation, followed by quantitative real-time PCR analysis of candidate genes.

A total of 273 differentially expressed genes were identified, including 27 DEEGs. Enrichment analysis revealed pathways related to neutrophil activation, NOD-like receptor signaling, and extracellular matrix organization. Six key genes, including S100A9, MMP9, FN1, NLRP3, CD55, and ITLN1 were selected as candidate diagnostic biomarkers. The multigene model demonstrated good diagnostic discrimination within the merged training cohort, while individual genes retained diagnostic value in an independent dataset. Several candidate compounds, including losartan, metoprolol, and pioglitazone, showed favorable binding affinities to key targets in molecular docking analyses. Transmission electron microscopy and nanoparticle tracking analysis confirmed the successful isolation of plasma exosomes, and qRT-PCR revealed significantly elevated expression of the six candidate genes in AMI-derived exosomes.

This integrative study identifies a set of exosome-associated genes with potential diagnostic relevance in AMI and provides exploratory in silico drug candidates targeting these biomarkers. These findings are hypothesis-generating and warrant further validation in large, prospectively collected cohorts and functional studies before clinical application.