Introduction

In this repository, I will describe our recently-released Structural Identifiability Toolbox, a web-based application for assessing parameter identifiability of differential models.

Click here to checkout the application! Read on to learn more.

Why is it better?

The program is fast, free, and is available in any web-browser, including mobile. We take care of a lot stuff in the background letting the user worry only about their model definition without any technicalities.

Who is it designed for?

Simply put, if you are working with models based on ordinary differential equations and you wish to know structural identifiability properties of your models' parameters then this app is built for you!

Knowing structural identifiability properties will help one set up better experiments to obtain data for the underlying process and extract parameter values correctly.

What can it do?

Individual Parameters

The application can answer questions about local or global identifiability parameters (including initial conditions): you provide the input system, choose the probability of correctness and, optionally, specify which parameter to check (by default it checks for all possible parameters). For this, the app is using SIAN algorithm¹² for details) which is fast, robust, and is correct with user-specified probability.

Parameter Combinations

If a parameter is non-identifiable, one may wish to seek an identifiable function that contains this parameter (and, possibly, others). Built with a brand new algorithm[^3], the app can quickly assess generators for all such functions. Moreover, we provide a way to assess whether 1 or more experiments are required to do so, this is called multi-experiment identifiability.

References

[3]: Computing All Identifiable Functions of ODE Models, arXiv:2004.07774