The present chapter provides a concise introduction to Bayesian learning of Bayesian networks, and then discusses two problems of particular relevance to systems biology: systematically integrating biological prior knowledge into the inference process, and learning regulatory networks from non-homogeneous temporal processes. We demonstrate the application of these methods on two network reconstruction problems: learning a protein signalling pathway from flow cytometry data, and learning a gene regulatory network related to circadian regulation in the model plant Arabidopsis from microarray gene expression time series.