CutTree performance. CutTree outperforms other search methods in gene network simulations, both in the number of experiments needed to find the PAGs and in the number of PAGs found. (a) Average number of experiments needed to find different numbers of PAGs. Average values from 50 in silico simulations; the random search bars are the expected values. (b) Comparison between CutTree and the Network identification method. The two Network identification results correspond to 90% and 50% coverage, that is the fraction of identified.