Optimally Designed Model Selection for Synthetic Biology.

Area of Science:

• Synthetic biology
• Computational biology
• Systems biology

Background:

• Automating the Design-Build-Test-Learn cycle in synthetic biology is hindered by challenges in modeling parts and circuits.
• Model discrimination requires significant data, computational resources, and expertise, often leading to suboptimal model choices.

Purpose of the Study:

• To outline frequentist and Bayesian approaches for model discrimination in synthetic biology.
• To demonstrate that optimally designed experiments can efficiently discriminate between competing models using a genetic toggle switch as a test case.

Main Methods:

• Application of frequentist and Bayesian frameworks for model discrimination.
• Ranking of three candidate models for a genetic toggle switch using in vivo data.
• Dynamical-systems interpretation and sensitivity analysis of optimization results.

Main Results:

• Efficient model discrimination was achieved using optimally designed experiments within both frequentist and Bayesian frameworks.
• Optimal experimental design effectively distinguishes predictions from competing models by exploiting specific input space regions.
• The study provides a dynamical-systems interpretation of optimization outcomes and their parameter sensitivity.

Conclusions:

• Optimal experimental design is a powerful strategy for discriminating between competing models of synthetic gene regulatory networks.
• This approach enhances the objectivity and efficiency of model selection in synthetic biology.
• Perturbations designed to maximally distinguish model predictions are key to effective discrimination.