Robust Moiety Model Selection Using Mass Spectrometry Measured Isotopologues

Area of Science:

• Metabolomics
• Systems Biology
• Computational Biology

Background:

• Stable isotope resolved metabolomics (SIRM) is crucial for metabolic flux analysis and modeling.
• Model correctness assumptions can significantly impact the interpretation of metabolic flux results.
• Existing methods may not adequately address model selection challenges in SIRM.

Purpose of the Study:

• To develop and validate a metabolic modeling software package for moiety model comparison and selection.
• To assess the effectiveness of model selection using time-series mass spectrometry isotopologue datasets.
• To investigate the influence of optimization parameters on model selection accuracy.

Main Methods:

• Developed a novel metabolic modeling software for moiety model comparison.
• Utilized two time-series mass spectrometry (MS) isotopologue datasets for UDP-GlcNAc.
• Tested model selection robustness across various optimization methods and criteria.

Main Results:

• Successfully selected the optimal model from over 40 candidates, demonstrating method robustness.
• Illustrated the impact of optimization methods, degree of optimization, and selection criteria on model selection.
• Showed that over-optimization can lead to model selection failure, while combining datasets mitigates overfitting.

Conclusions:

• The developed software effectively facilitates moiety model selection in SIRM experiments.
• Combining multiple SIRM datasets, including public repositories, improves model selection reliability.
• High data quality curation in public metabolomics repositories is vital for advancing metabolic modeling.