Bayesian nonparametric modelling with Gaussian processes in systems biology
Prof. D. Husmeier (Glasgow University)
Friday 23rd March, 2012 16:00-17:00 515
The proper functioning of any living cell relies on complex networks of gene regulation, and elucidating their structure and regulatory mechanisms is the principal objective of contemporary systems biology. Various modelling approaches based on nonlinear dynamical systems have been attempted. However, system identification, parameter inference and model comparison remain challenging tasks due to the sparsity of experimental replications, the nonlinear nature of the likelihood surface and the computational complexity of the learning algorithm. In my talk, I will describe an approach based on nonparametric Bayesian statistics with Gaussian processes, which allows accelerated inference without having to explicitly solve the underlying differential equations. I will describe the current state of the art, compare the method with related alternative paradigms, and conclude with a discussion of our own approaches. This is joint work with Frank Dondelinger and Ji Ruirui.