Estimating mutation rates in a Markov branching process using approximate Bayesian computation | JoVE Visualize

Area of Science:

Evolutionary Biology
Microbial Genetics
Computational Biology

Background:

Estimating microbial mutation rates is vital for understanding evolution, disease dynamics, and antibiotic resistance.
Traditional methods like Luria-Delbrück analysis rely on strict assumptions that often fail in real-world microbial experiments.
Deviations from model assumptions limit the applicability and accuracy of existing mutation rate estimators.

Purpose of the Study:

To develop a novel, robust method for estimating microbial mutation rates that overcomes the limitations of traditional approaches.
To introduce a likelihood-free computational framework for analyzing microbial mutation experiments, accommodating complex biological scenarios.
To enhance computational efficiency in mutation rate estimation through surrogate modeling.

Main Methods:

Modeling microbial mutation data using a two-type Markov branching process (MBP).
Employing approximate Bayesian computation (ABC) for parameter estimation, leveraging intensive simulations.
Integrating a Gaussian process (GP) surrogate model within the ABC framework to create the GPS-ABC estimator for improved efficiency.

Main Results:

ABC-based estimators, particularly GPS-ABC, demonstrate superior performance compared to traditional moment or likelihood-based estimators for constant mutation rates.
The GPS-ABC method effectively estimates mutation rates even in complex scenarios, such as two-stage mutations with piece-wise constant rates, where traditional methods fail.
The proposed GPS-ABC estimator was successfully applied to real experimental datasets for analyzing drug resistance in microbes.

Conclusions:

The proposed GPS-ABC method provides a powerful and flexible tool for accurate microbial mutation rate estimation, especially in biologically realistic, complex settings.
This likelihood-free approach enhances the analysis of microbial evolution, virology, and antibiotic resistance studies.
GPS-ABC represents a significant advancement in computational methods for microbial population genetics and experimental evolution.