Reverse-Engineering Biological Interaction Networks from Noisy Data using Regularized Least Squares and Instrumental Variables

The problem of reverse engineering the topology of a biological network from noisy time–series measurements is one of the most important challenges in the field of Systems Biology. In this work, we develop a new inference approach which combines the Regularized Least Squares (RLS) technique with a technique to avoid the introduction of bias and nonconsistency due to measurement noise in the estimation of the parameters in the standard Least Squares (LS) formulation, the Instrumental Variables (IV) method. We test our approach on a set of nonlinear in silico networks and show that the combined exploitation of RLS and IV methods improves the predictions with respect to other standard approaches.