Bayesian analysis of physiologically based toxicokinetic and toxicodynamic models

Area of Science:

• Toxicology and Environmental Health
• Computational Biology and Bioinformatics
• Risk Assessment and Modeling

Background:

• Physiologically based toxicokinetic (PBTK) and toxicodynamic (TD) models are crucial for estimating human toxic doses from animal studies, particularly for substances like bromate.
• Accurate calibration of these complex, highly parameterized models is essential for reliable internal dose predictions.
• Traditional frequentist methods (e.g., MLE, least squared error) struggle with parameter uncertainty and variability in complex models.

Purpose of the Study:

• To describe the Bayesian approach and Markov chain Monte Carlo (MCMC) analysis for calibrating complex PBTK/TD models.
• To explain the application of MCMC in risk assessment for improving dose estimations.
• To outline the advantages and challenges associated with using MCMC in toxicological modeling.

Main Methods:

• Utilized Bayesian inference, incorporating prior biological knowledge as parameter distributions.
• Employed Markov chain Monte Carlo (MCMC) simulation to refine parameter estimates using experimental data, generating posterior distributions.
• Focused on calibrating PBTK/TD models for bromate to improve accuracy in dose prediction.

Main Results:

• Demonstrated that MCMC successfully calibrates highly parameterized toxicokinetic models, overcoming limitations of frequentist approaches.
• Showcased how prior knowledge and experimental data are integrated to produce robust posterior parameter estimates.
• Facilitated more accurate estimation of internal doses and associated uncertainties for risk assessment.

Conclusions:

• The Bayesian MCMC approach offers a powerful and flexible method for calibrating complex toxicokinetic models.
• This technique enhances the reliability of human health risk assessments by improving the accuracy of dose estimations derived from animal data.
• Understanding MCMC is valuable for toxicologists and risk assessors seeking to leverage advanced computational methods.