First level Bayesian network, leukemia data. This diagram shows the most-likely network as derived from the frequency and stability of the network nodes found during cross-validation. All features show high mutual information with the class variable. The first level features are those which cannot be made independent of the class when conditioned on other features. Features placed on the second (lowest) level below the class have little mutual information with the class when conditioned on the parent feature.
Kuschner et al. BMC Bioinformatics 2010 11:177 doi:10.1186/1471-2105-11-177