A Probabilistic Framework for Molecular Network Structure Inference by Means of Mechanistic Modeling

Area of Science:

• Systems Biology
• Computational Biology
• Molecular Systems Engineering

Background:

• Ordinary differential equations (ODEs) are vital for mechanistic modeling of molecular networks.
• Inferring ODE model structure from time-course data is challenging, especially with numerous potential mechanisms.
• Current methods often fail to capture uncertainty in network component selection.

Purpose of the Study:

• To develop a novel Bayesian approach for model structure inference in ODE models.
• To enable probabilistic inference of molecular network structures, accounting for uncertainty.
• To apply the method to biological data for inferring gene regulatory networks.

Main Methods:

• A novel Markov chain Monte Carlo (MCMC) approach was formulated.
• A Metropolis algorithm was employed for efficient exploration of model space.
• The method was validated using simulated data and applied to time-course RNA sequencing data.

Main Results:

• The MCMC approach allows for probabilistic inference of ODE model structures.
• The method was successfully applied to infer the regulatory network of T helper 17 (Th17) cell differentiation.
• Results align with known three-phase differentiation process and provide phase-specific interaction predictions.

Conclusions:

• The developed Bayesian MCMC method offers a robust framework for ODE model structure inference.
• This approach enhances understanding of molecular network dynamics and component contributions.
• The study provides new insights into the regulatory mechanisms of Th17 cell differentiation.