De-composing the composite network into condition specific regulatory links. To predict the experimental conditions for each TF-target gene link in the regulatory network, we unified the set of experimental conditions generated by the LINK and STAR models. We applied both models to the local networks containing the TF→Gene1 link. (a) A set of experimental conditions (A) in which the TF→Gene1 link is predicted to be active according to the LINK model. Under conditions A, Gene1 and TF are over-expressed. (b) A set of experimental conditions (B) in which the targets are activated by TF according to the STAR model. Under conditions B, Genes1,2,3 are stimulated and Gene4 is suppressed by TF. (c) Finally, we determine the experimental conditions in which the TF→Gene1 link is active by taking the union of A and B. The quality of our network de-composition has been assessed by using the union A ∪ B.