(A) Regulatory network elements or building blocks studied. The element types are classified into those involving only target-target relationships and those involving only regulator-target relationships. For each type of element, a diagram illustrating a typical element structure is shown. The black triangles are transcriptional regulators, and yellow boxes are transcription factor target genes. The red and green arrows correspond to repressing and activating regulatory interactions respectively. The consistency measure used to evaluate the agreement between a particular element and gene expression data is also indicated. Feed-forward loops are not considered to be a basic building block of the network, but are included here because ignoring them could lead to overestimating the consistency of regulatory interactions. (B) A schematic illustration of the strategy used to calculate the statistical significance of a particular value of a consistency measure (in this case Pearson correlation coefficient for a regulatory interaction). A set of random network elements corresponding to the particular network element studied is created, and the same consistency measure is calculated for each random element. The resulting distribution of values is used to evaluate the statistical significance of the true value of a consistency measure in the form of a P-value. The P-value is computed as the fraction of random elements with the squared consistency measure value higher than the squared observed value for the actual element. This corresponds to the estimated probability of observing a squared consistency measure value as high as the one for the actual element for a regulatory network element with random regulator and target genes.

Summary of the results of the consistency calculations for yeast. See the caption of Figure 2 for explanations of the legends and abbreviations. (A) Results for regulator modules. The numbers of regulator-target pairs are: All (907), Act (643), and Rep (264). (B) Results for multiregulator modules. The numbers of target genes in multiregulator modules with different numbers of regulators are: All (269), 1 (155), 2 (73), 3 (33), and 4 (8). (C) Results for regulatory interactions. The numbers of interactions are: All (925), Act (651), and Rep (274). (D) Results for target modules. The numbers of target genes with different numbers of regulators are: All (413), 1 (173), 2 (113), 3 (72), and 4 (23).

Maps of the yeast regulatory network indicating the magnitudes of the consistency metrics and consistent network elements at a P-value threshold of 0.01. See caption for Figure 4 for details. (A) Network maps with each gene labeled by its name. The subcomponents identified for the yeast network are (A, amino acid utilization; C, carbon utilization; D, drug response; E, cell cycle control; I, nitrogen utilization; N, nucleotide utilization; O, oxygen response; P, phosphate utilization; S, stress response). (B) Consistent target-target relationships. (C) Consistent regulator-target relationships.

Summary of the results of the consistency calculations for Escherichia coli. The results are presented in the form of fraction of consistent network elements for three different significance levels (P < 0.01, P < 0.005, P < 0.001). (A) Results for regulator modules classified by the mode of regulation of the regulator acting on the module. All, all modules with more than one target gene (number of regulator-target pairs n = 1355); Act, activator-regulated modules (844); Rep, repressor-regulated modules (455); Dual, dual function regulator-regulated modules (61). (B) Results for multiregulator modules shown for all modules as well as modules with different numbers of regulators (n_reg). The results are only shown for cases with >10 instances of the module with the indicated number of regulators. The total number of target genes in multiregulator modules with more than one target is 658, and the numbers of target genes in multiregulator modules with different numbers of regulators are (in format: number of regulators [number of targets with this number of regulators]): 1 (368), 2 (142), 3 (71), 4 (46), and 5 (23). (C) Results for regulatory interactions classified by the mode of the regulator. The numbers of regulatory interactions in the network are: All (1367), Act (855), Rep (451), and Dual (61). In addition to the P-value threshold, results are also shown for consistency criteria utilizing both P-value threshold and sign criteria (s) as described in the text for pairwise correlations. (D) Results for target modules classified by the number of regulators acting on a target gene. The results are only shown for cases with >10 instances of the module with the indicated number of regulators. The numbers of target modules with different numbers of regulators are (in format: number of regulators [number of targets]): 1 (386), 2 (168), 3 (77), 4 (54), and 5 (25).

Maps of the Escherichia coli regulatory network indicating the magnitudes of the consistency metrics and consistent network elements at a P-value threshold of 0.01. (A) Network maps with each gene labeled by its name. Triangles correspond to regulators and boxes to targets. Particular subcomponents of the network are identified by labeled boxes (C, carbon utilization; D, DNA metabolism; F, flagellar biosynthesis; I, iron utilization; N, DNA damage; O, osmotic stress; P, phosphate utilization; R, purine utilization; S, stress response; T, Trp/Tyr utilization; U, sulfur utilization). (B) Consistent target-target relationships in the regulatory network. The widths of the links are proportional to R_RM² indicating how coherent the target gene expression is with the expression of other genes in the regulator module defined by the regulator involved in the interaction. For target genes (and regulators that are also targets of other regulators) the sizes of the nodes are proportional to R_MRM² indicating how coherent the target gene expression is with the expression of other genes in the same multiregulator module. The colors of the links indicate the mode of regulation (green, activation; red, repression; blue, dual). The darker shades of colors indicate that this component of the regulator module (for links) or multiregulator module (for nodes) is consistent with gene expression data (at P < 0.01). (C) Consistent regulator-target relationships in the regulatory network. The widths of the links are proportional to R_RI² indicating the correlation between regulator and target expression levels. The sizes of the nodes are proportional to R_TM² and describe how well the target expression level is explained by the regulator expression levels. The color scheme is the same as in B. The sign criterion was used to evaluate the consistency of regulatory interactions as described in the text.