Learning Gene Networks with Bayesian Networks
This chapter discusses several common problems and applications of Bayesian networks. The field of Bayesian networks emerged as a very promising one in computational systems biology soon after the introduction of microarray technology, as the technology provided the data needed to determine the existence and type of dependencies between genes in gene networks. The Bayesian Belief Network model provides an appealing alternative to the source-based approach. Bayesian networks are useful when most variables are independent of each other, in other words, when the dependency graph is sparse. Probabilistic inference with Bayesian networks uses the prior knowledge together with evidence regarding some of the network’s variables to produce probabilistic statements regarding the value of other variables in the network. In gene networks the expression profiles of all pairs can be compared using measures of statistical similarity or independence, such as correlation or mutual information.