A Comparison of Network Reconstruction Methods for Chemical Reaction Networks

Chemical reaction networks model biological interactions that regulate the functional properties of a cell; these networks characterize the chemical pathways that result in a particular phenotype. One goal of systems biology is to understand the structure of these networks given concentration measurements of various species in the system. Previous work has shown that this network reconstruction problem is fundamentally impossible, even for simplified linear models, unless a particular experiment design is followed. Nevertheless, reconstruction algorithms have been developed that attempt to approximate a solution using sparsity or similar heuristics. This work compares, in silico, the results of three of these methods in situations where the necessary experiment design has been followed, and it illustrates the degradation of each method as increasing noise levels are added to the data.