Bayesian Generalized Linear Model for Simulating Bacterial Inactivation/Growth Considering Variability and...

Area of Science:

• Microbiology
• Statistical Modeling
• Bioinformatics

Background:

• Conventional regression analysis (least-squares method) uses a normal distribution, neglecting bacterial variability and uncertainty.
• Accurate modeling of bacterial kinetics is crucial for food safety and risk assessment.

Purpose of the Study:

• To propose and validate a Bayesian generalized linear model (GLM) for bacterial population dynamics.
• To incorporate variability and uncertainty into bacterial kinetic modeling.
• To enhance the accuracy of bacterial inactivation and growth predictions.

Main Methods:

• Developed a Bayesian GLM incorporating Poisson and negative binomial distributions for residuals.
• Applied the model to inactivation kinetics of *Bacillus simplex* and growth kinetics of *Listeria monocytogenes*.
• Validated the model using over 50 replications with varying initial cell counts.

Main Results:

• The Bayesian GLM accurately described bacterial inactivation and growth data across multiple experimental runs.
• Over 90% of observed cell numbers fell within the 95% prediction interval, demonstrating high model accuracy.
• Parameter uncertainty was quantified and expressed as probability distributions.

Conclusions:

• Bayesian GLM effectively captures bacterial population variability and uncertainty.
• The model provides a robust framework for analyzing bacterial kinetics and improving foodborne pathogen risk assessment.
• This approach offers consistent analysis procedures applicable to simulation and fitting processes.