Data-driven analysis and prediction of dynamic postprandial metabolic response to multiple dietary challenges using...

Area of Science:

• Metabolomics
• Systems Biology
• Precision Nutrition

Background:

• Predicting individual metabolic responses to diet is crucial for personalized nutrition.
• Current analysis tools struggle with complex data from crossover studies.
• Existing methods lack predictive power or require extensive mechanistic knowledge.

Purpose of the Study:

• To apply dynamic mode decomposition (DMD), a data-driven method, to time-resolved metabolomics data.
• To integrate parametric DMD (pDMD) and DMD with control (DMDc) for enhanced analysis.
• To predict dynamic metabolic responses and identify inter-individual differences (metabotypes).

Main Methods:

• Utilized parametric DMD with control (pDMDc) on time-resolved metabolomics data.
• Applied the method to a crossover study with repeated measures.
• Leveraged a data-driven approach to model linear dynamical systems.

Main Results:

• Successfully predicted metabolite responses to unseen dietary exposures (R²=0.40 for measured data).
• Demonstrated predictive capability on simulated data (r=0.65).
• Identified clusters of dynamic metabolite responses, revealing metabotypes.

Conclusions:

• pDMDc is a powerful tool for analyzing time-resolved metabolomics data in precision nutrition.
• The method enables prediction of dietary metabolic responses and identification of individual differences.
• pDMDc shows potential for guiding metabolite responses toward target levels in personalized nutrition.