Bioinformatics and machine learning-based exploration of psychological stress-related genes: clinical significance for diagnosis and molecular classification in acute myocardial infarction

Abstract

Acute myocardial infarction (AMI) substantially contributes to global mortality and disability, and psychological stress constitutes a crucial correlative factor in its pathogenesis. This study aimed to identify potential biomarkers linked to psychological‑stress‑related genes (PSRGs) in patients with AMI. Four integrated gene‑expression training datasets were obtained from the Gene Expression Omnibus (GEO) database, while PSRGs were retrieved from the GeneCards database. Multiple machine‑learning algorithms were utilized to screen feature genes, and ROC curves were used to assess diagnostic performance. Gene‑set enrichment analysis, ssGSEA and CIBERSORT were applied to explore signaling pathways and immune‑infiltration patterns. A diagnostic signature containing seven PSRGs (CTSD, ADM, SOCS3, CCL5, TSPO, STAT1, and S100A8) was constructed, attaining a cross‑validation AUC of 0.933 within the training cohort. GSE66360 was used for secondary candidate‑gene refinement based on expression consistency and signature reproducibility evaluation (refitted‑model AUC = 0.957). AMI patients were classified into molecular subtypes with distinct immunological traits according to 49 differentially expressed PSRGs. RT‑qPCR verified differential expression of these signature genes between AMI and control samples. Collectively, our findings provide correlational transcriptomic insights into PSRG‑associated AMI pathogenesis and support the development of candidate diagnostic biomarkers and molecular subtyping strategies.