Integrative machine learning and Mendelian randomization identify causal laboratory biomarkers for coronary artery...

Area of Science:

• Pediatric Cardiology
• Artificial Intelligence in Medicine
• Genetics

Background:

• Kawasaki disease (KD) poses a significant risk of coronary artery lesions (CALs) in children.
• Early identification of high-risk patients is crucial for timely intervention and improved outcomes.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) model for predicting CAL risk in KD patients.
• To identify key predictive features and their causal associations with CALs.

Main Methods:

• Utilized machine learning (ML) models, including Multi-Layer Perceptron Classifier (MLPC), Random Forest (RF), and Extra Tree (ET), on data from 506 KD patients.
• Performed time-across validation and Mendelian randomization (MR) analysis to assess model performance and genetic causality.
• Identified seven predictive features for model development.

Main Results:

• MLPC demonstrated the highest accuracy among the validated ML models.
• Activated partial thromboplastin time (APTT) and red cell distribution width (RDW) were identified as causal factors associated with CALs.
• The AI model, supported by genetic validation, offers reliable risk stratification.

Conclusions:

• AI-based prediction models, integrated with genetic insights from MR, provide a robust method for early CAL risk assessment in KD.
• This strategy can facilitate prompt clinical management and potentially reduce the incidence of severe complications.