Development of Machine Learning methods for the kinetic modelling of metabolic networks for biotechnological and biomedical applications

Research in biological sciences is in the midst of a true revolution as a result of the emergence and growing impact of a series of new '-omics' disciplines and tools. These include genomics, transcriptomics, proteomics and metabolomics devoted respectively to the examination of the entire systems of genes, transcripts, proteins and metabolites present in a given cell or tissue type. This calls upon development of methods and tools that consolidate these disparate datasets to better understand the complex interactions of these biological processes at a systems level. Kinetic models of metabolism show the most potential in achieving this goal because of their ability to encapsulate all major cellular components and depict their complex life sustaining interactions in a quantitative and deterministic fashion. However, the primary bottleneck in building large-scale kinetic models of metabolism that accurately depict physiological and experimental observations is the uncertainties in the system parameters due to lack of consistent experimental data. In this thesis, we attempt to address this challenge by leveraging data driven approaches like generative machine learning to expedite the process of building kinetic models through two novel pipelines, REKINDLE and RENAISSANCE. We showcase the efficiency, versatility and scalability of these methods through a variety of studies addressing different behaviours of metabolic networks offering key insights at various scales. We anticipate that the frameworks will be convincing for the broader community to appreciate the utility of these models for biomedical and biotechnological purposes and also enable proper adoption of these models.