Bayesian Reverse Engineering of Biological Networks
Centre of Excellence for Mathematics and Statistics of Complex Systems Seminar
by Professor Michael Stumpf
Abstract: For most biological systems we lack reliable parameter estimates or well grounded mathematical models. In an ideal world we would be able to learn such models from sufficient amounts of data, and graphical models have are increasingly employed to learn aspects of the basic model structure.
Inference or reverse engineering of dynamical systems is more challenging and recent research into so-called sloppy parameters suggests that some aspects of dynamical systems may be nearly impossible to learn. Here we describe two complementary approaches to assess dynamical features of complex biological systems: (i) time-varying Dynamical Bayesian Networks
(tvDBN) infer the temporally changing structure of biological networks;
(ii) approximate Bayesian computation allows us to combine statistical inference for (stochastic and deterministic) dynamical systems with an assessment of inferability, sensitivity and robustness of model structures and parameters, respectively. We illustrate this Bayesian perspective on reverse engineering by application to several biological signalling processes
For More Information: Contact: Kostya Borovkov firstname.lastname@example.org