Related Experiment Video
Updated: Aug 11, 2026
08:10
Estimating Bilateral Atrial Function by Cardiovascular Magnetic Resonance Feature Tracking in Patients with Paroxysmal Atrial Fibrillation
Published on: July 20, 2022
Targeted metabolomic profiling and machine learning-based prediction models for persistent atrial fibrillation: a
Hao Gao1,2,3, Yanxiu Chen1,2,3, Pinliang Liao1,2,3
1Department of Cardiovascular Medicine, Southwest Hospital, Army Medical University, Chongqing, China.
Peerj
|August 10, 2026
Summary
Machine learning identified eight metabolites linked to persistent atrial fibrillation (PerAF) progression from paroxysmal atrial fibrillation (PAF). A simplified model accurately predicts PerAF risk, aiding clinical treatment adjustments.
Area of Science:
- Cardiology
- Metabolomics
- Machine Learning
Background:
- Atrial fibrillation (AF) is a common arrhythmia with high complication risks.
- Paroxysmal atrial fibrillation (PAF) can progress to persistent atrial fibrillation (PerAF).
- Early identification of PerAF progression is crucial for timely treatment adjustments.
Purpose of the Study:
- To identify metabolites associated with PerAF using machine learning (ML).
- To develop predictive models for the progression of PAF to PerAF.
- To enable personalized treatment strategies for AF patients.
Main Methods:
- Targeted metabolomic profiling of plasma from 100 AF patients (35 PAF, 65 PerAF).
- Univariate and multivariate analyses to identify differential metabolites and clinical features.
- Development and evaluation of four ML models (logistic regression, random forest, XGBoost, LightGBM) for PerAF prediction.
Main Results:
- Eight metabolites (e.g., kynurenine, citramalic acid) and two clinical features (NT-proBNP, uric acid) were significantly associated with PerAF.
- XGBoost model achieved high predictive performance (AUC discovery: 0.751, validation: 0.985).
- A simplified three-feature model (NT-proBNP, citramalic acid, uric acid) demonstrated robust predictive capability.
Conclusions:
- Identified novel PerAF-related metabolites and developed accurate ML-based predictive models.
- A simplified clinical model offers feasible risk stratification for PerAF progression.
- Further validation and exploration of metabolic pathways are warranted.