Reversible jump MCMC for multi-model inference in Metabolic Flux Analysis

Area of Science:

• Systems Biology
• Metabolic Flux Analysis (MFA)
• Computational Biology

Background:

• Model-based inference in systems biology relies heavily on model formulation.
• Numerous models with varying assumptions often exist for biological phenomena.
• Bayesian Model Averaging (BMA) addresses this by enabling inferences from multiple models but faces computational challenges with high-dimensional, non-linear models.

Purpose of the Study:

• To apply Bayesian Model Averaging (BMA) to Metabolic Flux Analysis (MFA) for inferring reaction directionality.
• To address the computational challenges of BMA in complex, high-dimensional biological models.
• To extend the scope of 13C MFA beyond parameter estimation to structural inference.

Main Methods:

• Utilized BMA within the framework of Metabolic Flux Analysis (MFA).
• Employed a tailored multi-model Markov Chain Monte Carlo (MCMC) approach.
• Applied the method to a large set of candidate models with different reaction directionality settings using 13C labeling data.

Main Results:

• Successfully inferred the in vivo probability of reaction bidirectionalities in a realistic metabolic network.
• Demonstrated the applicability of the developed algorithm in a complex setting.
• Extended the capabilities of 13C MFA to structural inference.

Conclusions:

• The developed BMA-based MCMC approach is effective for inferring reaction directionality in MFA.
• This method enhances the analysis of metabolic networks by providing structural insights.
• The study expands the utility of 13C MFA for understanding biological systems.