Application, Computation, and Theory for Synthetic Gene Circuits

Abstract

The field of synthetic gene circuits is concerned with engineering novel gene expression dynamics into organisms. This field, a subset of synthetic biology, was started almost two decades ago with the creation of two synthetic circuits: a bistable toggle switch and an oscillator. From the very outset, modeling has played a role in the development of synthetic circuits. However, modeling has been used to gain qualitative insight into dynamics, and actual quantitative modeling has been lagging behind. Parameters for quantitative models for biological systems often cannot be adequately estimated from measured data, because far too many sets of parameters can produce the same set of limited measured outputs. Additionally, models for synthetic gene circuits are often not correct the first time, and iterating on cycles of modeling and parameter estimation is difficult. Finally, there is a gap between development of modeling and system identification tools and their application to experiments on actual synthetic gene circuits. This thesis attempts to work towards addressing these issues with quantitative modeling for synthetic gene circuits. First, we derive theoretical conditions for identifiability of stochastic linear systems from heterogenous types of measurement data. These identifiability conditions can provide insight into what type of model to use and what measurements to collect in order to ensure that the resulting model can be identified. Second, we develop a software package for fast and flexible modeling and parameter estimation for synthetic gene circuits. The user can input models into our software using a simple text format and perform simulations of all types at optimized speeds. By inputting measured experimental data along with the model, the software can be used to perform Bayesian parameter estimation in an automated manner. To bridge the gap between computation and application, we apply this software to parameter estimation of DNA recombinase dynamics using real experimental data collected in an in vitro cell extract. Finally, we use modeling to guide the design of an improved single gene synthetic oscillator. While the original synthetic genetic oscillator contained three genes, we show that a simple circuit with a single gene can produce robust and synchronized oscillations across a population. Our results constitute an additional step towards the incorporation of quantitative modeling and parameter inference as part of the design-build-test cycle for synthetic gene circuits.