Abstract
<title>Abstract</title> <p> Background Ischemic cardiomyopathy due to myocardial infarction frequently induces adverse events. However, the lack of predictive models impedes clinical decisions. Objective We aimed to construct a gene-based model to predict post–myocardial infarction adverse cardiac events by integrating gene expression profiles and clinical factors. Methods This multicenter prospective cohort study was conducted between 2021 and 2024 and included patients from Jiangsu Provincial People’s Hospital and Hangzhou First People’s Hospital admitted between June 2021 and January 2024. All participants underwent follow-up until 36 months post-discharge. They were divided into model-construction (n = 291) and validation (n = 100) cohorts. The primary endpoints were major adverse cardiovascular events, including cardiac death, recurrent myocardial infarction, stroke, heart failure, or rehospitalization for further intervention. We used Cox regression analysis to identify risk factors. Model specificity and sensitivity were assessed via the concordance index, calibration curves, and receiver operating characteristic curves. Model performance was evaluated against previous models using integrated discrimination improvement and continuous net reclassification improvement. Kaplan-Meier curves were used for visualization and log-rank tests for between-group comparisons. Results Patients in the construction cohort (mean age: 63 ± 13 years) were predominantly male (77.7%). Among 55 patients who reached an endpoint event, 13 experienced cardiovascular death, 20 were readmitted for heart failure, and 22 underwent further intervention for chest pain. In the validation cohort, 25 patients experienced major adverse cardiovascular events. Biometric analysis and machine learning identified four candidate genes— <italic>FADS2</italic> , <italic>FMN1</italic> , <italic>TMEM176A</italic> , and <italic>RPS4Y1</italic> . A prognostic model combining polymerase chain reaction results and clinical factors achieved a concordance index of 0.83, increasing integrated discrimination improvement and continuous net reclassification improvement by at least 0.239 and 0.69, respectively, compared to previous models. The internal and external validation concordance indices were 0.864 and 0.723, respectively. Conclusions The developed prognostic model for myocardial infarction demonstrated robust sensitivity and specificity and favorable validation in external cohorts, offering substantial clinical guidance. Future studies should investigate the function and mechanism of action of identified targets. </p>