Optimizing regulation functions in gene network identification

This paper is concerned with the problem of identifying a discrete-time dynamical system model for a gene regulatory network with unknown topology using time series gene expression data. The topology of such a network can be characterized by a set of regulation hypotheses, one for each gene. In our earlier work, we formulated a convex optimization method to select the regulation hypotheses (and hence the network topology). In this paper, we further optimize the dynamics of the inferred network. Specifically, for a given topology, we minimize the ℓ₂ distance between the experimental data and the model prediction. We illustrate the performance of our algorithm by identifying models for gene networks with known topology.