Prediction of regulatory motifs by analysis of upstream sequences. (a) The average number of patterns discovered per gene group and (b) the average fraction (%) of the genes groups in the dataset in which at least one pattern was discovered are plotted as a function of the statistical significance of the patterns (Sig) (see Materials and methods for detail). The five plots displayed in each panel represent results obtained for the different datasets analyzed. These comprise three datasets of multiprotein complexes: the curated complexes from the CYGD catalogue , and the complexes identified by the TAP  and HMS  genome-scale analyses (see Materials and methods). The number of gene groups in these sets is 113, 124 and 269, respectively. The remaining two plots represent results obtained for the 79 annotated regulons (used as positive control), and for groups of genes of the same size as the considered complex or regulon, randomly selected from the yeast genome (used as negative control, as described in the text). Only regulons or complexes containing at least of five genes/proteins were considered.