Machine learning prediction of post-CABG atrial fibrillation using clinical and pharmacogenomic biomarkers

Lei Hua1, Jingxian Han1, Siqi Zhang1

  • 1Henan Key Laboratory of Cardiac Remodeling and Transplantation, The 7th People's Hospital of Zhengzhou, Zhengzhou, Henan, P.R. China.

Frontiers in Medicine
|September 29, 2025
PubMed

Insights

A new model predicts postoperative atrial fibrillation (POAF) risk after coronary artery bypass grafting (CABG) using clinical and genetic factors. This tool aids personalized perioperative care for better patient outcomes.

Area of Science:

  • Cardiology
  • Genetics
  • Artificial Intelligence

Background:

  • Postoperative atrial fibrillation (POAF) is a common complication after coronary artery bypass grafting (CABG).
  • POAF significantly affects patient prognosis and increases healthcare costs.
  • Accurate risk stratification is crucial for optimizing clinical management.

Purpose of the Study:

  • To develop an integrated predictive model for POAF risk stratification.
  • To optimize clinical management and personalized perioperative care for CABG patients.

Main Methods:

  • Retrospective analysis of 576 CABG patients from a cohort of 2,528 undergoing 21-gene pharmacogenetic testing.
  • Training and validation of eight machine learning algorithms using clinical variables and genetic variants.
  • Independent validation on a separate cohort of 61 patients.

Main Results:

  • The Gaussian Naive Bayes (GNB) model achieved high accuracy (0.81 in test set, 0.79 in validation set).
  • Key predictors identified include multivessel CABG, history of heart failure, rs5219 (KCNJ11), and prolonged bypass duration.
  • A web-based tool was developed for real-time POAF risk stratification.

Conclusions:

  • The GNB classifier integrates pharmacogenomic and clinical predictors for POAF risk assessment post-CABG.
  • The model serves as a valuable clinical decision-support tool.
  • This approach enhances personalized perioperative care through rigorous validation and user-centered design.
Abstract