Structural Identifiability Analysis via Extended Observability and Decomposition

Journal article

Structural identifiability analysis of nonlinear dynamic models requires symbolic manipulations, whose computational cost rises very fast with problem size. This hampers the application of these techniques to the large models which are increasingly common in systems biology. Here we present a method to assess parametric identifiability based on the framework of nonlinear observability. Essentially, our method considers model parameters as particular cases of state variables with zero dynamics, and evaluates structural identifiability by calculating the rank of a generalized observability-identifiability matrix. If a model is unidentifiable as a whole, the method determines the identifiability of its individual parameters. For models whose size or complexity prevents the direct application of this procedure, an optimization approach is used to decompose them into tractable subsystems. We demonstrate the feasibility of this approach by applying it to three well-known case studies.

Authors: Villaverde, A; Barreiro, A; Papachristodoulou, A

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Elsevier
• Journal: IFAC-PapersOnLine
• Volume: 49
• Issue: 26
• Pages: 171-177
• Publication date: 2017-01-06
• Acceptance date: 2017-01-01
• DOI: 10.1016/j.ifacol.2016.12.121
• ISSN: 2405-8963
• Keywords: systems biology; optimization; observability; parameter identification; model decomposition; structural identifiability