by Marwan, W., Wagler, A., Weismantel, R..
Series: 2006-10, Preprints
The reconstruction of biochemical and genetic networks from experimental data by generating realistic models of structure and function with predictive power is an important challenge in the post-genomic era. Here we describe an exact mathematical approach to the automatic reconstruction of regulatory networks from experimental data that reflect the time-dependent response to perturbation. Based on a mathematical proof, the procedure yields a complete list of all alternative network structures together with the minimal number of yet undetected network components required to reproduce a given data set. Starting with experimental data arrayed in a table, the generation and combinatorial processing of difference vectors finally yields the incidence matrix of a stochastic place-transition Petri net that accounts for the time-dependent signal flux. The procedure is outlined step by step by reconstructing a minimal network underlying a simple physiological response of a living cell to sensory stimulation. We subsequently show how Petri nets obtained by automatic reconstruction can be used to derive stochastic or deterministic quantitative models of biochemical or genetic networks that in turn can be used to design experiments aimed to specifically discriminate between members of a complete list of alternative network structures that account for the behavior of a system.
The automatic reconstruction procedure can also be used to modify a given network or to design a network through obtaining a provably complete list of all network structures that can exhibit a predefined behavior, in terms of inputs, outputs and the time dependent activity of any of its components.
network reconstruction, reverse engeneering