Performance of reverse-engineering for varying network sizes and experimental settings.

For each of the twelve combinations of size and experimental setting, 300 random reference networks were created. For each reference, a wild-type time-series and a varying number of knockout perturbations were simulated. A) Number of genes in networks. B) Number of different random single knockout experiments simulated. C) Number of different random double knockout experiments simulated. D) Percentage of cases where a predicted network was identical to the reference. E) AUPRC of the ensemble of all networks (mean/standard deviation). F) Percentage of cases where characteristic interaction sets have been identified. G) Number of runs where 2/3/4/5 characteristic sets were identified. More than 5 sets were not observed.