In this work, we study the piecewise linear class of models for the formulation of gene interactions. We assume that mRNA concentrations are determined by piecewise linear differential equations. An algorithm to determine a local state transition matrix and a statistial rule to estimate the switching conditions of the local model are presented. We show that the model class can be used for inference, and it can handle essential qualitative features of genomic regulation. First, we will introduce a hybrid system formulation of the gene network. Then, we will describe an algorithm to infer the locally linear system parameters exploiting the special structure of Boolean networks.