ABC SMC for parameter estimation and model selection with applications in systems biology

Abstract

AbstractApproximate Bayesian Computation (ABC) methods can be used in situations where the evaluation of the likelihood is computationally prohibitive. They are thus ideally suited for analyzing the complex dynamical models encountered in systems biology, where knowledge of the full (approximate) posterior is often essential.This talk gives an overview of an ABC algorithm based on Sequential Monte Carlo (ABC SMC). Different uses of the algorithm will be presented, depending on the application question of interest. The first is the general parameter estimation framework, where the interest lies in estimating the posterior parameter distribution from available experimental data. In the second context we ask whether the model can reproduce a desired qualitative or semi-quantitative behavior, and what dynamic behaviors the system can achieve. The third context discussed here for use of the ABC SMC algorithm is that of model selection. Here we ask which of the models (i.e. model topologies) from a pool of proposed candidate models represents the most suitable hypothesis about the biological system of interest.The focus of the presentation is applications of ABC SMC to questions from systems biology. ABC SMC is applied to a variety of biological models, including stochastic and deterministic descriptions of eukaryotic signaling pathways and prokaryotic stress response pathways. In particular, the algorithm is employed to understand understand the qualitative behavior of the phage shock protein response in bacteria Escherichia coli, and the model selection algorithm is applied to distinguish between differential equation models of MAP kinase phosphorylation dynamics and the JAK-STAT signaling pathway.