Uncertainty quantification, propagation and characterization by Bayesian analysis combined with global sensitivity...

Area of Science:

• Computational Biology
• Systems Biology
• Mathematical Modeling

Background:

• Intracellular dynamical models grow in complexity with experimental data.
• Over-parameterization leads to uncertainty in parameter estimates and model predictions.
• Need for methods to refine models and guide experimental design.

Purpose of the Study:

• To combine Bayesian analysis with global sensitivity analysis (GSA) for improved model predictions.
• To identify model weaknesses and parameters crucial for distinguishing qualitative behaviors.
• To guide future experimental design and model building efforts.

Main Methods:

• Utilized approximate Bayesian computation (ABC) for parameter estimation and uncertainty quantification.
• Propagated parameter uncertainty to model predictions (forward uncertainty propagation).
• Performed GSA on predictions using posterior distributions from ABC.

Main Results:

• Identified parameters with the largest contribution to prediction uncertainty.
• Pinpointed parameters essential for differentiating between qualitatively distinct model outputs.
• Applied methodology to a model of synaptic plasticity using multi-source experimental data.

Conclusions:

• The combined Bayesian analysis and GSA approach effectively addresses model over-parameterization.
• This methodology enhances prediction accuracy and provides critical insights for experimental design.
• The approach is valuable for refining complex biological models and understanding parameter influence.