Inferring the dynamics of yeast cell cycle control using Bayesian statistics

Abstract

Mathematical mechanistic modelling can be a useful tool to understand the dynamics of biological systems. However, due to the nested uncertainties and limited and noisy data, it is hard to develop good models. Bayesian statistics handles these difficulties naturally, but it introduces a new, computational problem: calculating high dimensional posterior probabilities and marginal likelihoods. In the context of dynamical models, several numerical approximation methods have previously been used. In this Master thesis, after evaluating & selecting the most powerful of these methods a specific variant of MCMC sampling it is applied to the study of yeast cell cycle control. Using Bayes factors to select between models, it was possible to evaluate different hypotheses regarding a negative feedback cycle in yeast cell cycle control. Stretching this approach further by integrating diverse datasets reveals that approximation methods which can more efficiently explore rugged and high-dimensional probability distributions are needed.