Simulation and optimization tools to study design principles of biological networks

Abstract

Recent studies have developed preliminary wiring diagrams for a number of important biological networks. However, the design principles governing the construction and operation of these networks remain mostly unknown. To discover design principles in these networks, we investigated and developed a set of computational tools described below. First, we looked into the application of optimization techniques to explore network topology, parameterization, or both, and to evaluate relative fitness of networks operational strategies. In particular, we studied the ability of an enzymatic cycle to produce dynamic properties such as responsiveness and transient noise filtering. We discovered that non-linearity of the enzymatic cycle allows more effective filtering of transient noise. Furthermore, we found that networks with multiple activation steps, despite being less responsive, are better in filtering transient noise. Second, we explored a method to construct compact models of signal transduction networks based on a protein-domain network representation. This method generates models whose number of species, in the worst case, scales quadratically to the number of protein-domain sites and modification states, a tremendous saving over the combinatorial scaling in the more standard mass-action model was estimated to consist of more that 10⁷ species and was too large to simulate; however, a simplified model consists of only 132 state variables and produced intuitive behavior. The resulting models were utilized to study the roles of a scaffold protein and of a shared binding domain to pathway functions.