Artificial intelligence and biomarker-driven prediction of post-coronary artery bypass grafting atrial fibrillation: Integrating clinical, genomic, and metabolic insights.

Background: Postoperative atrial fibrillation (POAF) remains one of the most frequent and consequential complications after coronary artery bypass grafting, occurring in 15%-30% of patients and contributing to increased morbidity, prolonged hospitalization, and higher long-term mortality. Despite decades of investigation, traditional risk models have shown limited predictive accuracy owing to the multifactorial nature of POAF. Objective: This review synthesizes the emerging evidence from recent studies applying artificial intelligence (AI) and machine learning (ML) approaches for POAF prediction, with focus on clinical, biochemical, genomic, and molecular dimensions. Methods: We conducted a structured narrative review of PubMed/MEDLINE, Embase, and Google Scholar for studies evaluating AI and ML-based prediction of POAF following isolated coronary artery bypass grafting published from January 2020 through 2025. Search terms included POAF, coronary artery bypass grafting, ML, AI, pharmacogenomics, and biomarkers. A total of 9 studies met inclusion criteria and were synthesized narratively given the heterogeneity in study designs and outcomes. Results: Across studies, modern algorithms demonstrate areas under the curve receiver operating characteristic between 0.80 and 0.93, with performance exceeding that reported for traditional clinical risk scores. However, external validation, model calibration, and biases remain to be fully addressed before clinical translation is considered. Conclusion: The convergence of interpretable AI, metabolic and genetic biomarkers, and clinical data offers a promising path toward individualized risk stratification and targeted postoperative management.