Figure 1

(Color) Partial target inhibition strategies. Panel (A), complete knockout: complete inhibition of a single target modeled by the elimination of all interactions at the representing node. Panel (B1), partial knockout: partial inactivation of the target by knocking out half of its interactions. Panel (B2), attenuation: partial inactivation of the target by attenuating the interactions of the representing node to 50% as an average. Panel (C1), distributed knockout: inactivation of individual interactions between nodes. Panel (C2), distributed attenuation: attenuation of individual interactions between nodes. In the attenuation experiments, attacking an edge at one end resulted in a 50% weakening of the interaction (dashed line). If a subsequent attack is directed against the other end of the edge, the interaction is weakened to 25% (dotted line).Reuse & Permissions

Figure 2

(Color) Effect of single-target and various multitarget attack strategies on network efficiency. The effect of a series of successive attacks is shown on the network efficiency (NE [25]; see Methods) of the regulatory networks of E. coli [21] or S. cerevisiae [22]. Each attack point was chosen to produce the maximal possible damage to the system. Panels (A) and (B), single-target attack was performed by eliminating all the edges of a given node [blue; cf. Fig. 1a]; partial knockout was modeled by fully blocking (removing) a randomly chosen half of the edges belonging to a given node as shown in Fig. 1, panel (B1). This attack was applied simultaneously to 2 (red), 5 (green), and 10 (black) nodes. Panels (C) and (D), attenuation was modeled by decreasing the contribution of edges belonging to a given node as shown in Fig. 1, panel (B2). The color codes are the same as in panels (A) and (B). Distributed systemwide knockout was modeled by either fully blocking [removing, panel (E)] or attenuation [panel (F)] an edge so as to produce a maximum decrease in NE, as schematically shown in Fig. 1, panels (C1) and (C2), respectively. In the attenuation experiments an edge could be attenuated at both ends; i.e., the maximal attenuation of a single edge was fourfold (from the initial 100% to 25%). For this reason the number of attacks [panel (F)] and the number of edges affected (Table , column 12) do not necessarily coincide. Blue and red signs of panels (C) and (D) refer to data from E. coli and S. cerevisiae, respectively.Reuse & Permissions

Figure 3

(Color) Sites affected by the various strategies in the E. coli (A) and the S. cerevisiae (B) regulatory networks. The attacks were carried out with the maximum damage algorithm based on the rigorous search strategy described in Methods. The strategies are those defined in Fig. 1, and the nodes and edges are the same as those described in Table .Reuse & Permissions