Machine learning-driven integrative analysis identifies an inflammation-related metabolic gene signature for acute myocardial infarction

Background Acute myocardial infarction (AMI) is a leading cause of global mortality, closely associated with inflammation and metabolic disturbances. This study aimed to develop a predictive model based on the inflammation-related metabolic genes (IRMGs) for the diagnosis and risk assessment of AMI, and to explore the upstream regulatory patterns and molecular interactions of the model’s core genes. Methods Gene expression datasets GSE66360 and GSE59867 were sourced from the Gene Expression Omnibus (GEO) database. Differentially expressed inflammation-related metabolic genes (DE-IRMGs) were identified from GSE66360. Candidate biomarkers for AMI were screened using LASSO regression, Random Forest, and extreme gradient boosting (XGBoost) algorithms. Stratified 20-fold cross-validation was applied to evaluate model stability after feature selection. A diagnostic model based on these biomarkers was constructed, followed by the prediction of target miRNAs, transcription factors (TFs), and potential therapeutic agents. The diagnostic performance of the model was validated using the external GSE59867 dataset and our clinical cohort. Immune infiltration profiles were analyzed using CIBERSORT and single-sample gene set enrichment analysis (ssGSEA). Finally, the mRNA expression levels of the identified biomarkers were verified through quantitative real-time polymerase chain reaction (RT-qPCR). Results A diagnostic signature for AMI was established based on five DE-IRMGs, namely PDE4B, ACSL1, NAMPT, ALDH2, and GLUL. This five-gene signature exhibited strong discriminatory performance in the training discovery cohort with 20-fold internal cross-validation (AUC = 0.938). When evaluated in independent cohorts, this fixed-coefficient model showed moderate discrimination in the external transcriptomic cohort GSE59867 (AUC = 0.685), while exploratory RT-qPCR analysis conducted in our in-house clinical cohort achieved an AUC of 0.972. Additionally, immune cell infiltration analysis revealed significant correlations of GLUL, ALDH2, NAMPT, and ACSL1 with various innate immune cell populations, including neutrophils, monocytes, macrophages, activated mast cells, and activated natural killer (NK) cells. Finally, RT-qPCR verification confirmed the significantly differential expression of these signature genes between AMI patients and control subjects. Conclusion The five‑gene signature shows potential as an exploratory diagnostic biomarker for AMI, offering insights for the early diagnosis and risk assessment of AMI.