A New Computational Tool for Establishing Model Parameter Identifiability

Abstract

Abstract

We describe a novel method to establish a priori whether the parameters of a nonlinear dynamical system are identifiable—that is, whether they can be deduced from output data (experimental observations). This is an important question as usually identifiability is assumed, and parameters are sought without first establishing whether these can be inferred from a set of measurements. We highlight the connections between parameter identifiability and state observability. We show how observability criteria can be used to check for identifiability, and we use new, state of the art computational tools to implement our approach. Nonlinear dynamical systems are prevalent in systems biology, where they are often used to represent a biological system. Thus, examples from biology are used to illustrate our method.