An interpretable machine learning model for precise prediction of biomarkers for intermittent fasting pattern.

Area of Science:

• Metabolomics
• Biomarker Discovery
• Machine Learning in Nutrition

Background:

• Intermittent fasting (IF) is a popular dietary approach with growing interest in its physiological effects.
• Identifying reliable biomarkers is crucial for understanding the metabolic impact of IF.
• Fecal metabolites offer a window into gut health and systemic responses to diet.

Purpose of the Study:

• To identify potential fecal biomarkers distinguishing intermittent fasting from normal feeding patterns.
• To apply interpretable machine learning for biomarker discovery in metabolomics data.
• To provide a foundation for healthier dietary lifestyle choices.

Main Methods:

• Untargeted metabolomics analysis of fecal samples from mice using Ultra-Performance Liquid Chromatography-High-Resolution Mass Spectrometry (UPLC-HRMS).
• Development and evaluation of five machine learning models for pattern recognition.
• Application of Shapely Additive Explanations (SHAP) for weighted biomarker analysis.

Main Results:

• The Random Forest model demonstrated the highest accuracy in differentiating between intermittent fasting and normal feeding groups.
• Shapely Additive Explanations (SHAP) analysis identified specific metabolite contributions.
• Ganoderenic Acid C was identified as a potential key biomarker for distinguishing the two dietary patterns.

Conclusions:

• Ganoderenic Acid C is a promising fecal biomarker for intermittent fasting.
• Interpretable machine learning, particularly Random Forest with SHAP, is effective for biomarker discovery in dietary studies.
• This research offers insights into metabolic responses to IF and supports evidence-based dietary recommendations.