Quantifying uncertainty in dose-response screenings of nanoparticles: a Bayesian data analysis.

Area of Science:

• Environmental Science
• Toxicology
• Computational Biology

Background:

• Nanoparticle risk assessment relies on dose-response data.
• Existing methods often lack robust uncertainty quantification.
• Accurate hazard metrics are crucial for effective risk management.

Purpose of the Study:

• To develop a Bayesian model for analyzing nanoparticle cytotoxicity dose-response data.
• To incorporate multiple sources of uncertainty, including assay sensitivity and historical data.
• To optimize experimental design for cost- and time-efficient toxicity screenings.

Main Methods:

• A Bayesian approach using a log-logistic model for zinc oxide (ZnO) nanoparticle cytotoxicity data.
• Accounting for unequal variances across doses and assay sensitivity (resazurin assay).
• Integrating historical data to complement experimental findings.

Main Results:

• Probability distributions for toxicity potency (EC50) and slope (s) were determined, providing uncertainty measures via credibility intervals.
• Benchmark dose (BMD) upper and lower limits were calculated with 95% probability.
• The Bayesian model identified optimal experimental designs for minimizing data while reducing uncertainty.

Conclusions:

• The Bayesian framework offers a comprehensive approach to nanoparticle risk assessment.
• It effectively quantifies uncertainty in hazard metric estimation.
• This methodology supports more efficient and reliable toxicity testing, particularly for ZnO nanoparticles.