A primary goal of systems biology is to integrate large bodies of experimental biological data together to make a predictive model. However this goal begs two questions. First how can we best merge disparate sources of data such as from expression arrays, proteomics analyses, and cell behavior assays into a coherent model? Second what future experiments best test the accuracy a systems level model? One framework well suited to answer both questions is a Bayesian network. A Bayesian network is a probabilistic, machine-learning algorithm that has found widespread use in business and engineering due to its flexibility and noise tolerance. In this talk I will show how Bayesian networks provide a rational route to combine data from previous experiments plus predict which future experiments will yield the most insight at the systems level.