Reconstruction of Gene Regulatory Networks using an Error Filtering Learning Scheme

One of the fundamental and most challenging problems in system biology is the reconstruction of gene regulatory networks from input-output data based on non-linear differential equations. This paper presents an approach to estimate the unknown nonlinearities and to identify the true network that generated the data, based on an error filtering learning scheme and a Lyapunov synthesis method. Unknown nonlinearities are modelled by networks using radial basis functions and model validation is performed by taking advantage of the so-called persistency of excitation of input signals, a condition that is shown to play a significant role in the problem of uncovering the true network structure. The proposed methodology and the theoretical results are validated through an illustrative example.