Benchmarking tools for a priori identifiability analysis

TL;DR

This paper evaluates 12 software tools for a priori structural identifiability analysis across 25 case studies, providing insights into their performance, strengths, and weaknesses to guide future tool development.

Contribution

It offers a comprehensive benchmarking of existing tools for a priori identifiability analysis, highlighting their computational efficiency and limitations.

Findings

Tools vary significantly in computational performance.

Some tools are more suitable for large models.

Guidelines for selecting appropriate tools are provided.

Abstract

The structural identifiability and the observability of a model determine the possibility of inferring its parameters and states by observing its outputs. These properties should be analysed before attempting to calibrate a model. Unfortunately, such \textit{a priori} analysis can be challenging, since it requires symbolic calculations that often have a high computational cost. In recent years a number of software tools have been developed for this task, mostly in the systems biology community but also in other disciplines. These tools have vastly different features and capabilities, and a critical assessment of their performance is still lacking. Here we present a comprehensive study of the computational resources available for analysing structural identifiability. We consider 12 software tools developed in 7 programming languages (Matlab, Maple, Mathematica, Julia, Python, Reduce, and Maxima), and evaluate their performance using a set of 25 case studies created from 21 models. Our results reveal their strengths and weaknesses, provide guidelines for choosing the most appropriate tool for a given problem, and highlight opportunities for future developments.