AbstractCreating organisms with predetermined traits has long been a goal of synthetic biology and has the potential enable the creation of cell factories, medical tools and bio-materials. The genome is a key factor in controlling attributes of an organism and biologists have made great progress in writing and editing genomes, ushering in an era of synthetic organisms. However, limited understanding of how phenotypes emerge from genotypes has contributed towards a lack of genome design tools leaving genome engineers unsure what genome edits to perform or genomes to create. Computer models hold great potential in aiding the design of genomes but the tools are designed for specific models and do not adapt well. Additionally, the models focus on specific processes in a cell (e.g. metabolism) and miss control mechanisms that feed into those processes from other processes (e.g. gene regulation of metabolic enzymes). Whole-cell modelling may enable models that can incorporate systems-level effects and minimal genomes may simplify the genotype phenotype relationship. Unfortunately, there are currently no methods to accurately find minimal genomes and there is only one published whole-cell model that is hard to use and too computationally expensive to use for large-scale in-silico experiments.
This thesis aims to create tools to design in-silico genomes by enabling massive in-silico experiments on state-of-the-art computational models. These tools should be easily adaptable to different models, computers clusters, design goals, and design algorithms. As a proof-of-concept these tools will be used to try and reduce the genome of the only published whole-cell model. The results of this will be further analysed in the hope of learning about the whole solution space of genome reductions and how it may help genome reduction algorithms.
Towards these goals we present the genome design suite which enables massive in-silico experiments across multiple computer clusters and can avoid maximum simulation times imposed on the cluster. The code is designed to make it easily adaptable. The genome design suite was used to perform over 100,000 in-silico gene knockout experiments to develop genome reduction algorithms and reduce the whole-cell model of M. genitalium by 165 genes. This minimal genome is described biologically and further analysis of the simulations reveal multiple paths of convergence to the minimal genome, high and low-essential gene combinations and the role of dynamic gene essentiality in shaping the solution space.
We conclude that the genome design suite can aid in-silico genome design. The ability to avoid maximum simulation times on clusters and utilise multiple clusters enables larger-scale in-silico experiments giving new insights into solution spaces. Its adaptability allows it to evolve to new models, design goals, and design algorithms. It also enables the genome design tools built on it to utilise new models quickly and vice versa.
|Date of Award||24 Mar 2020|
|Supervisor||Lucia Marucci (Supervisor) & Claire S Grierson (Supervisor)|