Global Parametric Identifiability of a Dynamic Unstructured Model of Hybridoma Cell Culture

Abstract The question of identifiability of a dynamic model is deciding, based on the mathematical behaviour of the model itself - before supporting experimental data have been gathered and analysed - whether all of the parameters within the model can, in principle, be uniquely identified from data. In this paper, we propose a new dynamic optimisation-based identifiability test for the determination of global parametric identifiability of nonlinear dynamic process models. We provide a description of the optimisation framework, detailing a semi-infinite programming (SIP) problem whose solution determines identifiability, and, through inspection of the parameter values obtained, identifies the problematic parameters. For cases in which time-varying inputs to the process are allowed, the problem is cast as an optimal control problem. Within this framework, one can calculate optimal fixed and variable external controls with input constraints, and initial conditions of a dynamic experiment. The new approach can be shown to overcome model size and form limitations often encountered in past approaches. We demonstrate the use of the methods on a dynamic unstructured model of hybridoma cell culture.