Cell-fate decision model: in silico training sets. Training sets used to evaluate the optimization method. Each training set is given in the form of a transition graph, where each node contains a perturbation with the corresponding observation (stable phenotype) and edges denote a transition between stable phenotype upon change of condition (perturbation). Perturbations are specified using a combination of node names prefixed by â orâ +â sign to specify knock-out (node state fixed to 0) or over-expression (node state fixed to 1). When multiple edges connect the same source node to different target nodes, it means that all transitions start from the same phenotype (same attractor of the network). Node states that were intentionally reversed to generate errors in training sets are shown in red. (PDF 9015 kb)