Correlation of expression patterns and presence of regulatory motifs. Histograms representing the distribution of pairwise expression correlations for all pairs of genes within a co-regulated set, as compared to the distribution among all pairs of differentially expressed genes (regardless of putative regulation). The expression correlation c(i, j) between two genes i and j is the uncentered correlation metric used for clustering of gene expression in Figure 5A (i.e., the cosine of the angle between the expression vectors for genes i and j). Distributions are (red) over all pairs of genes within a specified set, or (blue) over all 386 genes in the full differentially expressed set. (A) HrpL; (B) PvdS; (C) Fur; (D) RpoD. The sizable spike at c = 0 (i.e., representing uncorrelated gene pairs) in the background distribution results from the sparse nature of the data array: for a gene that is not deemed to exhibit significant differential expression in a given condition, its fold change is set to zero for the purposes of clustering and computing expression correlations.