An efficient moments-based inference method for within-host bacterial infection dynamics.

Area of Science:

• Microbiology
• Computational Biology
• Biostatistics

Background:

• Isogenic tagging (IT) has advanced bacterial infection studies in animal models.
• Quantitative analysis of IT data is limited by underdeveloped statistical models.
• Stochastic Markovian models show promise for bacterial population dynamics.

Purpose of the Study:

• To present an efficient numerical method for fitting stochastic dynamic models to in vivo experimental IT data.
• To overcome the limitations of simulation-based inference in complex bacterial infection models.
• To enable accurate parameter estimation for bacterial population dynamics.

Main Methods:

• Solving systems of ordinary differential equations for the first two moments (mean, variance, covariance).
• Estimating model parameters using divergence minimization for models with linear dynamic rates.
• Comparing results with maximum likelihood estimation using an experimental dataset.

Main Results:

• The new method efficiently and accurately estimates model parameters.
• Parameter estimates showed overlapping confidence regions with maximum likelihood estimation.
• The novel method yielded lower bacterial division and death rates, improving goodness-of-fit at later time points.

Conclusions:

• The developed numerical method provides an efficient framework for analyzing bacterial infection dynamics from IT data.
• This computational efficiency facilitates model comparison and optimal experimental design.
• The flexible framework is applicable to diverse experimental systems in microbial research.