Intelligent Engineering of Bacterial Genomes

The ambition of this project is to establish a generic methodology for the intelligent engineering of biological hosts for the production of high-value drugs and chemicals through synthetic biology.
The project combines three research areas pioneered by the supervising team, in a unique interdisciplinary approach: advanced genome editing; predictive computational modelling; and statistical learning strategies. It exploits capabilities in high-throughput robotics, large-scale screening and machine learning available in the participating groups.
Strain optimization is by far the most time-consuming and expensive bottleneck in the pathway to commercialization of any compound produced by engineered microbes. An intelligent integrated approach to designing "pre-fabricated" host strains for the high-level overproduction of industrially relevant chemical classes would make a major contribution to overcoming this bottleneck and accelerating the way to market for a large number of biotechnology applications.
The application case targeted in this project focuses on creating an optimized Escherichia coli host for the efficient production of compounds of interest (e.g., antimicrobials and anticancer agents). The work will include three closely interlinked work packages, which will interact through an iterative design - build - test - learn cycle in several rounds during the lifetime of the project.
Work package 1: Establish a computational pipeline for the design of strain engineering approaches, combining both the assembly and integration of libraries of biosynthetic pathways towards the compounds of interest and the modelling of E. coli metabolism, incorporating resource allocation constraints and ensemble modelling strategies.
Work package 2: Model-driven disruption using genome editing. This work package will use available high-quality genome-scale models of E. coli metabolism to predict a set of enzyme-coding genes as targets for disruption or overexpression. Engineered strains will be realized in a modular design of experiments, using high-throughput genome editing, and the resulting strain collection will be rapidly phenotyped by metabolomics and targeted analytics in a panel of relevant growth conditions.
Work package 3: Statistical learning to optimize genome editing strategies. We will employ machine learning approaches developed in collaboration with the industrial partner, Cambridge Consultants Ltd., to identify successful editing strategies (taking into account metabolic and regulatory interactions) and guide the next iteration of mutagenesis, strain design and genome editing.
The project provides comprehensive interdisciplinary training at the interface of biology, engineering and bioinformatics, essential for the next generation of biotechnology scientists.