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

Bayesian methods are advantageous for biological modeling studies due totheir ability to quantify and characterize posterior variability in model parameters.When Bayesian methods cannot be applied, due either to nondeterminismin the model or limitations on system observability, approximateBayesian computation (ABC) methods can be used to similar effect, despiteproducing inflated estimates of the true posterior variance. Due to generallydiffering application domains, there are few studies comparing Bayesian andABC methods, and thus there is little understanding of the properties andmagnitude of this uncertainty inflation. To address this problem, we presenttwo popular strategies for ABC sampling that we have adapted to performexact Bayesian inference, and compare them on several model problems. Wefind that one sampler was impractical for exact inference due to its sensitivityto a key normalizing constant, and additionally highlight sensitivities ofboth samplers to various algorithmic parameters and model conditions. Weconclude with a study of the O’Hara-Rudy cardiac action potential modelto quantify the uncertainty amplification resulting from employing ABC usinga set of clinically relevant biomarkers. We hope that this work servesto guide the implementation and comparative assessment of Bayesian andABC sampling techniques in biological models