Figure 10.

Success rate of retrieving missing links from simulated data. In each trial, two random (metabolite, regulator) links are added to the joint network, and their functions are randomly assigned. The maximum path length is 6. The feedback links are counted as accurately learned if they are among the degenerate link sets generated by the algorithm. The success rate is the fraction of random trials which accurately learn the feedback links from simulated data. The success rate increases as the number of perturbation experiments increases.