Bayesian inference informed by parameter subset selection for a minimal PBPK brain model.

Area of Science:

• Pharmacokinetics and Systems Biology
• Computational Biology
• Mathematical Modeling

Background:

• Physiologically based pharmacokinetic (PBPK) models simulate drug absorption, distribution, metabolism, and excretion (ADME) using ordinary differential equations.
• These complex models require accurate parameter quantification and uncertainty analysis for reliable clinical application.
• Parameter identifiability, determining which parameters are uniquely defined by data, is crucial for robust PBPK modeling.

Purpose of the Study:

• To develop and validate a strategy for integrating parameter subset selection with Bayesian inference in PBPK models.
• To enhance the accuracy of PBPK model predictions through refined parameter estimation and uncertainty quantification.
• To reduce the computational cost associated with complex PBPK model analyses.

Main Methods:

• Implementing a strategy that combines parameter identifiability analysis with Bayesian inference.
• Utilizing frequentist or Bayesian inference methods for uncertainty quantification.
• Applying parameter subset selection to refine the set of identifiable parameters within PBPK models.

Main Results:

• Successful integration of parameter identifiability analysis and Bayesian inference for PBPK models.
• Demonstrated refinement of identifiable parameter subsets, leading to improved model predictions.
• Quantification of both parameter and response uncertainties, enhancing model reliability.
• Reduction in computational demands for PBPK model analysis.

Conclusions:

• The proposed strategy effectively refines PBPK models by focusing on identifiable parameters.
• Integrating identifiability analysis with Bayesian inference enhances prediction accuracy and quantifies uncertainty.
• This approach offers a computationally efficient method for robust PBPK model development and application in healthcare.