Figure 2

Procedures of the genetic algorithm. Ms (M1, M2, ...) indicate motifs. A series of motifs (e.g., |M1|M7|M5|M2|) is a motif combination composed of the motifs. At the initialization step, motifs are randomly selected to generate motif combinations. For each motif combination, matching scores are calculated from upstream sequences, and expression levels of mRNAs with the upstream sequences are obtained from microarray data. Regression between the scores and the expression levels is performed to obtain the goodness-of-fit (AIC or GCV), and motif combinations with the best goodness-of-fit values are selected to take crossover. Then mutation is performed to replace motifs with other motifs. These procedures are iteratively executed for updated motif combinations.