Fig. 5: Illustrative example of categorizing perturbation experiments and scoring model agreement.

A The reported results are normalized by the activation of the response node R in the absence of the signal (S = 0). The wild type shows higher activation of R in the presence of the signal S. M and C represent mediators of the signal. ”KO” means knockout and “CA” means constitutive activation. The wild type response under S = 0 is categorized as OFF; the wild type response under S = 1 is categorized as ON. If the value of R observed under a perturbation is similar to one of these two reference values, it will be included in the same category as the reference. For example, the response to [S = 1, C CA] is categorized as ON. Notably, the response to [S = 0, C CA] is categorized as Some. B Illustrative example of determining the minimal trap space agreement and the hierarchy-based score of putative models. Each model indicates the next state of R (denoted R*) as a function of the current state of S or M. The model with R* = 0 has the same minimal trap space agreement as the R* = S model, and in particular it agrees with the observation [S = 1, M KO] → R = 0, but it receives a lower hierarchy-based score because of its discrepancy with the observation [S = 1] → R = 1. In general, the difference between the minimal trap space agreement and the hierarchy-based score (highlighted in yellow) is more prominent in more complex perturbation experiments.