Trans-dimensional Bayesian model averaging for 13C-metabolic flux analysis: Evidence-based flux inference under...

Area of Science:

• Systems Biology
• Biotechnology
• Metabolic Engineering

Background:

• Accurate intracellular metabolic flux quantification is crucial for systems biology and biotechnology.
• 13C-metabolic flux analysis (MFA) is the standard for flux estimation but often suffers from insufficient data, leading to model-dependent results.
• Existing methods struggle to address structural uncertainty in metabolic network models.

Purpose of the Study:

• To develop a scalable Bayesian inference framework for 13C-MFA that accounts for uncertainty in metabolic network structures.
• To enable robust flux estimation by averaging over multiple possible network configurations.
• To quantify the support for alternative pathway hypotheses and propagate structural uncertainty into flux estimates.

Main Methods:

• Introduced Bayesian model set averaging, a scalable Bayesian inference framework for 13C-MFA.
• Combined reversible jump Markov chain Monte Carlo for exploring structural hypotheses with diffusive nested sampling for model evidence estimation.
• Applied the framework to synthetic case studies at illustrative and application scales.

Main Results:

• The method yields robust flux estimates even with structural uncertainty.
• It effectively identifies when multiple network configurations are statistically indistinguishable.
• The framework recovers correct, data-supported pathway configurations as data informativeness increases.

Conclusions:

• Bayesian model set averaging provides a practical foundation for quantitative Bayesian flux inference under structural model uncertainty in 13C-MFA.
• The approach scales to billions of model variants, offering a significant advancement over previous methods.
• This framework allows for a more comprehensive understanding of metabolic networks by explicitly addressing structural ambiguity.