Comparing two sequential Monte Carlo samplers for exact and approximate Bayesian inference on biological models.

Area of Science:

• Computational Biology
• Statistical Modeling
• Biophysics

Background:

• Bayesian methods excel in biological modeling by quantifying parameter variability.
• Approximate Bayesian computation (ABC) is used when Bayesian methods are infeasible, but inflates posterior variance.
• Limited research exists comparing Bayesian and ABC methods, leaving uncertainty inflation properties unclear.

Purpose of the Study:

• To adapt two popular ABC sampling strategies for exact Bayesian inference.
• To compare the performance of these adapted samplers on various model problems.
• To quantify uncertainty amplification in ABC methods using a cardiac action potential model.

Main Methods:

• Adapted two common ABC sampling strategies to perform exact Bayesian inference.
• Evaluated sampler performance across several biological model problems.
• Applied adapted samplers to the O'Hara-Rudy cardiac action potential model with clinical biomarkers.

Main Results:

• One adapted ABC sampler proved impractical for exact inference due to normalizing constant sensitivity.
• Both samplers exhibited sensitivities to algorithmic parameters and model conditions.
• Quantified significant uncertainty amplification when using ABC with clinical biomarkers in the cardiac model.

Conclusions:

• Adapted ABC samplers offer a pathway for exact Bayesian inference in biological modeling.
• Understanding and mitigating ABC's uncertainty inflation is crucial for reliable biological models.
• This work provides guidance for implementing and comparing Bayesian and ABC sampling techniques in biological research.