Fig. 3.

The stress-induced mutation network. (A) Protein-protein interactions: CytoScape 2.8.3 software, “unweighted force-directed layout” (28), links from STRING 9.0 (12). Proteins that promote σ^S, σ^E, and SOS activation (Fig. 4), as green, black circle, and red circle, constitute 54% of the network. Downstream of SOS (7), solid red. (B) Coexpression and protein-protein interaction are significantly more clustered than random controls. Gene expression data (13). The 93 SIM genes, (92 × 93)/2 = 4278 pairs, show correlation coefficient distributions (top): bars, entire range; boxes 25th and 75th percentile; red bars, mean. Of 4278 pairs, 3350 show positive correlation coefficient; 928 lie below the zero threshold level. High statistical significance for the strong phenotype (S) genes is increased by addition of moderate (M) and weak (W) (table S3). (Bottom) Significantly more protein-protein interactions for SIM than random genes. Of 4278 pairs, 1320 show positive interaction scores; 2958 pairs do not. P values: sign test of the probability of failure to reject the null hypothesis “number of positively correlated pairs is the same as in the random control.” (C) Allocation of network genes upstream of stress responses (data summarized in tables S1 and S7).