The authors demonstrate how to use graphical models, and Bayesian networks in particular, to model genetic regulatory networks. Bayesian network methods are well suited to this domain because of their ability to model more than pair-wise relationships between variables, their ability to guard against overfitting, and their robustness in the face of noisy data. The authors discuss how to combine genomic expression and binding location data to produce principled scores that allow models to be compared with one another directly. They develop methods for extending the semantics of Bayesian networks to include edge annotations for modeling statistical dependencies between biological factors with greater refinement. They also derive principled methods for scoring these annotated Bayesian networks in the presence of genomic data. They apply their scoring framework to validate models of regulatory networks in comparison with one another in the context of the Saccharomyces cerevisiae galactose regulatory system.