Modelling and measurement in synthetic biology

Abstract

Synthetic biology applies engineering principles to make progress in the study of complex biological phenomena. The aim is to develop understanding through the praxis of construction and design. The computational branch of this endeavour explicitly brings the tools of abstraction and modularity to bear. This thesis pursues two distinct lines of inquiry concerning the application of computational tools in the setting of synthetic biology. One thread traces a narrative through multi-paradigm computational simulations, interpretation of results, and quantification of biological order. The other develops computational infrastructure for describing, simulating and discovering, synthetic genetic circuits. The emergence of structure in biological organisms, morphogenesis, is critically important for understanding both normal and pathological development of tissues. Here, we focus on epithelial tissues because models of two dimensional cellular monolayers are computationally tractable. We use a vertex model that consists of a potential energy minimisation process interwoven with topological changes in the graph structure of the tissue. To make this interweaving precise, we define a language for propagators from which an unambiguous description of the simulation methodology can be constructed. The vertex model is then used to reproduce laboratory results of patterning in engineered mammalian cells. The assertion that the claim of reproduction is justified is based on a novel measure of structure on coloured graphs which we call path entropy. This measure is then extended to the setting of continuous regions and used to quantify the development of structure in house mouse (Mus musculus) embryos using three dimensional segmented anatomical models. While it is recognised that DNA can be considered a powerful computational environment, it is far from obvious how to program with nucleic acids. Using rule-based modelling of modular biological parts, we develop a method for discovering synthetic genetic programs that meet a specification provided by the user. This method rests on the concept of annotation as applied to rule-based programs. We begin with annotating rules and proceed to generating entire rule-based programs from annotations themselves. Building on those tools we describe an evolutionary algorithm for discovering genetic circuits from specifications provided in terms of probability distributions. This strategy provides a dual benefit: using stochastic simulation captures circuit behaviour at low copy numbers as well as complex properties such as oscillations, and using standard biological parts produces results that are implementable in the laboratory.