A systems and control engineering approach to synthetic biology.

View all posters

Kirubhakaran Krishnanathan, Stephen R.P. Jaffé, Sean R. Anderson, Phillip C. Wright, Stephen A. Billings and Visakan Kadirkamanathan

University of Sheffield, United Kingdom

A key challenge in synthetic biology is the characterisation of genetic parts. The need of obtaining kinetic data (time-series) in order to achieve this is undoubtedly important. A dynamic model which is able to interpret such data is an asset to characterisation. The cornerstone to the design of higher order genetic parts is a simplistic transcriptional-translational genetic part, which exhibits a switching characteristic that is nonlinear in dynamics. We take a systems level approach using only input/output data to model and analyse this behaviour (rather than a network based approach which takes into account the intermediate reactions taking place). By taking inspiration from the Nonlinear AutoRegressive Moving Average model (NARMAX), a concise and compact model which exhibits accurate representation of real data is derived. Genetic parts exhibit noise and variability which results in stochasticity and heterogeneity respectively, causing shortcomings in typical model performance. We address this and propose an identification framework using the NARMAX model and Approximate Bayesian Computation–Sequential Monte Carlo (ABC-SMC) technique. ABC-SMC provides the posterior distribution of the model parameters thereby capturing the observed variation in genetic parts. A quorum-sensing genetic part from the “registry of parts” is used as a case study to demonstrate the derivation of a parsimonious model whose parameters are estimated as a posterior distribution. The identified model is computationally convenient to analyse for design purposes, which is an advantage of the modelling approach.