Acute myocardial infarction (AMI) remains a leading cause of global cardiovascular morbidity and mortality. Limitations in current diagnostic methods hinder early detection and intervention, creating an urgent need for novel early diagnostic biomarkers. This study employed an integrated multi-omics approach, combining metabolomics, Mendelian randomization (MR), and transcriptomics data to identify potential AMI biomarkers. Plasma metabolomic profiling revealed 174 differentially abundant metabolites. Subsequent MR analysis pinpointed a key causal metabolite, L-arachidoyl carnitine (carnitine C20:0). Genes associated with this metabolite were retrieved from the GeneCards database and cross-referenced with differentially expressed genes from the GEO database, leading to the identification of 10 candidate biomarker genes: ACSL1, PYGL, DYSF, MGAM, SLC7A7, SULF2, KCNJ2, CYP1B1, NCF2, and SLC22A4. By constructing and evaluating 80 machine learning models, the Enet[alpha = 0.1] model was determined to have the optimal diagnostic performance. The diagnostic potential of these ten genes was further corroborated by logistic regression with tenfold cross-validation. Additionally, immune cell infiltration analysis using the CIBERSORT algorithm uncovered potential associations between the candidate genes and specific immune cell subpopulations. In conclusion, this sequential multi-omics investigation successfully identifies and validates 10 gene biomarkers related to AMI, offering new perspectives for early precision diagnosis and insights into the disease's pathogenesis, alongside potential therapeutic targets.
Keywords: Acute myocardial infarction; Diagnostic Markers; Machine Learning; Mendelian Randomization; Metabolomics.
© 2026. The Author(s).