Nested effects models. Markowetz et al.  introduce a probabilistic model to infer a pathway structure (left) from the observed downstream effects of interventions (right). The model predicts that genes high up in the pathway (like A) will have a broader set of effects than genes more downstream (like B to E). The branching in the pathway below B corresponds to (partly) disjoint effect subsets for C and D. The intersection of effect sets for C and D can be attributed to E and is explained by the collider at E in the pathway.