Systematic comparison of modeling fidelity levels and parameter inference settings applied to negative feedback gene regulation

Adrien Coulier, Prashant Singh, Marc Sturrock, Andreas Hellander

Abstract: Author summary Computational models plays a vital role in modern biology and are commonly used to compare theory with data. These models can take different forms, and there is often a trade-off between model detail and the computational resources needed to simulate them. Furthermore a choice must be made regarding how to compare the model output with the available data with several different distance metrics available. The choice of model and distance metric may also be impacted by the amount of available data. Therefore deciding how best to infer model parameters from available experimental data is a challenging problem. In this paper we have developed a computational pipeline designed to systematically evaluate inference accuracy for a wide range of true known parameters. To demonstrate its use, we applied it to a well studied gene regulation model. In particular, we compared a simple model, mid-complexity model and a complex model for several data-scenarios and for multiple numerical options for parameter inference. We believe this pipeline can be used as a preliminary step to guide scientists prior to gathering experimental data. This could prevent experimentalists from gathering unnecessary expensive experimental data or modelers from expending huge computational resources on simulating superfluously complex models.