Overview of GeneClass algorithm. (Left) Data representation for input to the GeneClass algorithm. P
represents the states of the parents π in each experiment e. M
represents the presence/absence of motifs μ in the promoter of genes g. y represents the label of each example with feature vector xg,e. (Right) The output of the algorithm is an alternating decision tree, which defines a prediction function for the up/down regulatory response of genes. Each splitter node represents a feature chosen by the algorithm, and the prediction node below it contains its prediction score.