## Example scenarios and optimizations for the example network in Figure 1.

Rows “sc1”, “sc2”, “sc3” correspond to scenarios 1 to 3. The “Perturbations” column shows the externally imposed state of the nodes and which can be −1 (downregulation), 0 (state of the node did not change), or 1 (activation level is increased). No value is given if the node was not perturbed. The “Measurements” column shows the measured change of the activation level of , and in the respective scenarios. The “Initial fitting error” column shows the total mismatch of predictions and measurements with respect to the initial topology (shown in Figure 1A). The “MCoS” (Minimal Correction Sets) column shows artificial positive (1) or negative (−1) external inputs to some nodes which would lead to a perfect fit of the data (resulting fitting error for the scenario becomes 0). The “Remaining fitting error” columns show the remaining mismatches for the optimal subgraph depicted in Figure 1B and for the two optimal graphs displayed in Figure 1C. The original network in Figure 1A has a total fitting error of 5; it is 2 for the optimal subgraph in Figure 1B and it becomes 0 in the optimal graphs in Figure 1C.