Fig. S1. The response space of the improved TOL model. An additional interaction for the repression of Pm was added to the TOL network. A total of seven response states were reached by simulating the revised TOL network by all possible input combinations. The order of the TOL protein is shown in the top and the relative frequency of each state in the 1024 simulations is shown at the bottom of each. On the right, the input dependences for each specific state, which are defined as the inputs that are required to generate the observed responses, are shown. The inputs that should be present for the given condition and those that should be absent are shown in green and red respectively. The level of dependence ranges from 1 (inputs found in all simulations) to −1 (inputs absent in all simulations). In addition, intermediate levels of dependence are allowed (i.e. variables required only in a specific set of conditions). Those variables that have a dependence of 0 are irrelevant in the given state.


Fig. S2. Analysis of TOL system robustness compared with 100 random networks. For the analysis, 50 perturbed copies of the TOL network were created and their attractors identified. Next, the percentage of original attractors found in the set of perturbed networks was measured (red vertical line) compared to the percentage of attractors found in 100 random networks, each also perturbed 50 times. The distribution of original attractors found in the random data set is shown as a histogram, the vertical blue line indicating the percentage of original attractors found in the 5% more robust systems. Note that the percentage of attractors found in the perturbed versions of the TOL network was higher than those found in > 99% of the randomized circuits (P-value of 0.01).

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.