Ensemble modelling strategies for the exploration of the gamma-butyrolactone network in Streptomyces coelicolor

Systems biology employs predictive models that guide hypothesis-driven research. As the use of computational models is increasingly gaining ground within the biological community and is moving from proof of concept to real-world applications, such as the engineering of synthetic biology circuits, there is a higher need for clear and testable predictions. However, parameter uncertainty is an important hindrance in the model analysis process. This thesis carefully considers the so called 'parameter problem' and suggests a framework that explicitly considers parameter uncertainty by sampling parameters from probability distributions. Thus, the pitfalls that traditional fitting strategies face can be avoided. By employing a 'respectful modelling' approach that rigorously quantifies the confidence associated with the modelling results and keeps all options in view rather than focusing on a single maximum likelihood solution, models can be made adaptable in the face of emerging experimental results. Another important consideration is the quality of the parameter distributions. As the issue of generating informative priors has not been successfully addressed yet, the thesis suggests a novel protocol that aims to fill this gap. The protocol concerns the collection of parameter values from a diverse range of sources (literature, databases and experiments), assessing their plausibility, and creating log-normal probability distributions, while maintaining the thermodynamic consistency of the model. Finally, the suggested methodological developments are put in action, in order to explore the gamma-butyrolactone regulatory system in Streptomyces coelicolor, a Gram-positive, soil bacteria. This small but complex system involves two genes (scbR and scbA) and regulates the antibiotic production through a mechanism of action that is not yet fully elucidated. The scenarios that have been suggested involve the formation of a putative ScbR-ScbA protein complex, potential transcriptional interference and antisense RNA interactions. The thesis describes the replication and ensemble modelling analysis of the previously published GBL models, as well as a model on the similarly structured quorum sensing (QS) system, in order to explore the behaviours of the two systems and allow comparisons between them. Additionally, a new, versatile and adaptable computational model was designed which considered all three suggested mechanisms in different combinations and with varying promoter strengths. The analysis of the model suggests that the most significant influence in the system's behaviour seems to come from antisense RNA interactions, combined with an aggressive scbR promoter. The model can be used to question and refine our understanding of the system's activity and could even suggest a different biological role than the one originally assumed. The model also indicates key experiments which could further elucidate the role of the system and the interactions of its components and ultimately lead to the design of robust and sensitive systems which can be used in synthetic biology and biotechnology.