Overview and details of the RIPE method.

<p>(A) Overview of the RIPE algorithm. (B) First step to obtain a large set of causal orderings from a network with cycles. The network graph is decomposed into strongly connected components, so called super-nodes, (left) followed by a DFS on the strongly connected components (right). For example, the post-visit time of super-node A is 8, and thus A precedes all other nodes. The topological ordering of super-nodes is ABCD. (C) Illustration of MC-DFS algorithm. Gene perturbation graph (left), DFS visit times for labeling #1 (middle), and DFS visit times for labeling #2 (right). (D) Depiction of a small network to illustrate the influence graph and the predictors used in the penalized likelihood estimation procedure. The true regulatory network (left), the influence graph under no noise (middle), observed influence graph, with false positive and false negative edges (right). Edges in the regulatory networks are shown in thick lines and additional edges in the influence graph are distinguished with narrow lines; red dash lines indicate false positives.