Nested sampling for Bayesian model comparison in the context of Salmonella disease dynamics

Area of Science:

• Computational Biology
• Systems Biology
• Statistical Modeling

Background:

• Comparing biological models traditionally uses maximum likelihood methods like Akaike's Information Criterion (AIC).
• Bayesian methods offer more informative posterior probability distributions but face computational challenges with large parameter spaces.
• Nested sampling is an efficient computational method for posterior probability calculation, previously applied in physical sciences.

Purpose of the Study:

• To demonstrate the application of nested sampling for inference and model comparison in biological sciences.
• To reanalyze Salmonella enterica infection data in mice using nested sampling.
• To provide a more comprehensive Bayesian analysis compared to traditional AIC methods.

Main Methods:

• Application of nested sampling for Bayesian model comparison.
• Reanalysis of experimental data on Salmonella enterica infection dynamics in mouse liver cells.
• Integration across parameter space and estimation of posterior parameter distributions.

Main Results:

• Nested sampling confirmed findings from the original AIC analysis.
• Provided integration across parameter space, enabling posterior parameter distribution estimation and visualization of correlations.
• Offered posterior predictive distributions for model goodness-of-fit assessments.

Conclusions:

• Nested sampling is a viable and informative method for biological model inference and comparison.
• Goodness-of-fit results suggest exploring alternative mechanistic models and relaxing quasi-stationary assumptions.
• The approach enhances understanding of complex biological system dynamics through robust statistical assessment.