Designing robust synthetic genetic circuits using approximate Bayesian computation

View all posters

Miriam Leon, Chris Barnes

University College London, United Kingdom

Creating synthetic devices that are robust to changing cellular contexts will be key to the success of synthetic biology. When faced with a set of competing designs for a given genetic circuit one is likely to choose the simplest possible model that can achieve the desired behaviour. However simple systems are often the least robust and it is well known that additional feedback interactions can increase robustness to extrinsic noise sources. Here we utilise a design methodology that takes advantage of Bayesian statistics. This allows us to use model selection to compare designs based on their robustness and handle uncertainty in biochemical rate constants. We examine small gene networks, implementing various common behaviours, which take an input signal and trigger a response – ultimately leading to the production of a protein as an output signal. For each network device, we consider multiple models, each one capable of generating the desired behaviour but containing different feedback connections. We consider both deterministic and stochastic processes for the dynamics and use Bayesian model selection to directly compare the competing designs for their ability to perform in the face of extrinsic noise. The method also provides distributions of the kinetic parameters required to give the desired input and output characteristics thus providing information on the strengths of the required feedbacks. Our results provide insight into strategies for constructing more robust synthetic devices.