Causal models and interventions

Florian Markowetz, CMB, MPIMG, Berlin.

Bayesian networks are directed graphical models. I will compare them to undirected (Gaussian) graphical models. From observational data, the reconstruction of network structure is limited to Markov equivalence classes. These can be resolved by using data from perturbation experiments (knock-outs, RNAi). I will introduce a score for Bayesian model selection designed especially for RNAi data.