A comprehensive multi-task deep learning approach for predicting metabolic syndrome with genetic, nutritional, and...

Area of Science:

• Genomics and Computational Biology
• Cardiovascular Disease Research
• Metabolic Health and Endocrinology

Background:

• Metabolic syndrome (MetS) is a cluster of conditions increasing cardiovascular disease and type 2 diabetes risk.
• Current prediction models treat MetS as binary, neglecting its five distinct components.
• Accurate, component-specific risk prediction is crucial for early intervention.

Purpose of the Study:

• To develop and evaluate a multi-task deep learning model for simultaneous prediction of MetS and its components.
• To compare the proposed model's performance against various single-task machine learning methods.
• To integrate genomic, lifestyle, dietary, and socio-economic data for comprehensive risk assessment.

Main Methods:

• Developed a multi-task deep learning framework to predict MetS and its five constituent factors.
• Utilized the Korean Association Resource (KARE) dataset, including 352,228 SNPs from 7,729 individuals.
• Incorporated genomic data alongside lifestyle, dietary, and socio-economic variables.

Main Results:

• The multi-task deep learning model demonstrated superior performance compared to single-task models.
• Evaluated using metrics including accuracy, precision, F1-score, and Area Under the ROC Curve (AUC).
• The proposed model effectively predicted MetS and its individual components with higher accuracy.

Conclusions:

• Multi-task deep learning offers a more effective approach for predicting MetS and its components.
• Integrating diverse data types enhances the predictive power for metabolic disorders.
• This model provides a promising tool for early identification and management of MetS risk.