Variability in responses to random perturbations constrain biomolecular network interactions

Perturbation experiments are a common technique to identify interactions in biochemical reaction networks; the response of the system of interest is observed after it is exposed to a range of drugs or disruptions. Average responses of cellular components are then analysed to infer the strengths of connections in the network. Previous approaches like Modular Response Analysis require specific knowledge of each perturbation's targets in the network and only analyse deterministic averages. We demonstrate that the correlation and variation of molecular responses can constrain the interaction topology between two connected components of interest, even when perturbation targets are unknown. In a model system we illustrate how this approach can distinguish between feedback through the other measured component and feedback that is independent of the other component. Finally, we quantify how well our approach—based on deterministic interactions and infinitesimal perturbations—describes biological systems with non-linear stochastic interactions and finite perturbations. This work can help analyse large-scale drug perturbation studies where the exact nature of each perturbation is unknown.