Performance of RIPE and competing methods on the reconstruction of synthetic networks.

(A) Average measures for reconstruction using NEM, PCALG, FFLDR and RIPE on a network of size . corresponds to the ideal influence graph and to represent increasing levels of noise in perturbation data (see also Figure S7). (B)–(C) Average measures for reconstruction of synthetic regulatory networks using PCALG, FFLDR and RIPE for different levels of false positive and negative noise in perturbation data. Numbers in parentheses indicate the expected number of false edges in each case. The true graph is an acyclic graph (DAG) of size in (B) and a cyclic graph of size in (C).