Identifying Key Features Associated with Excessive Fructose Intake: A Machine Learning Analysis of a Mexican Cohort.

Area of Science:

• Metabolomics
• Nutritional Science
• Data Science

Background:

• Excessive fructose consumption is associated with negative health effects.
• Limited research exists on clinical, behavioral, and nutritional patterns of fructose intake using machine learning.
• Understanding these patterns is crucial for public health.

Purpose of the Study:

• To apply machine learning algorithms to identify patterns associated with high fructose intake in a healthy Mexican cohort.
• To define high fructose consumption as intake exceeding 25 grams per day.
• To explore clinical, behavioral, and nutritional factors linked to fructose consumption levels.

Main Methods:

• Unsupervised (K-Means clustering) and supervised (Extreme Gradient Boosting, Random Forest, Histogram-based Gradient Boosting) machine learning algorithms were employed.
• Analysis was conducted on a healthy Mexican cohort.
• Shapley Additive Explanations (SHAPs) were used for model interpretation.

Main Results:

• K-Means clustering revealed three distinct subgroups, one with less favorable anthropometric, biochemical, and behavioral characteristics.
• Supervised models achieved balanced accuracies around 80% and Area Under the Curve (AUC) up to 88.1% in distinguishing intake levels.
• Key features associated with high fructose intake included body mass index, triglycerides, sleep duration, alcohol consumption, and anxiety indicators.

Conclusions:

• Fructose consumption is influenced by multiple factors, highlighting its complex nature.
• Machine learning effectively uncovers dietary and metabolic patterns related to fructose intake.
• These findings can inform future nutrition strategies and warrant further investigation.