New and improved methods for integrated models of metabolism and gene expression

The beginning of the 21st century was marked by the advent of disruptive technologies, which ushered an era of groundbreaking advances in fundamental sciences, carried by the great pace at which computational capabilities spread and evolved. But the new century also came with its fair share of challenges. Anthropogenic climate change brought issues of sustainability in the production chemicals and food. The global improvement of life expectancy and diagnostic methods also saw the increased incidence of illnesses for which age is a risk factor, such as cancer and dementia. Systems biology, and more specifically metabolic engineering, are fields that are interested in engineering living organisms. In particular, they provide tools and methods that are suited to respond to the new requirements of chemical production, food availability, and health and medicine through the understanding and engineering of living cells. Engineered microorganisms are already used in the production of both commodity and specialty chemicals, genetically improved crops are a possible answer to the ever-increasing food demand, and new medical treatments rely on an improved understanding and control of cellular idiosyncrasies. Efficient engineering requires mathematical models. Over the last decades, the increasing availability of full genome sequences and their translation into models of metabolism enabled the emergence of a wide gamut of methods to describe the inner workings of the cells we study. In particular, models of metabolism and gene expression (ME-models) were the first formulation to account simultaneously for cell metabolism, and the expression mechanisms translating genetic information into proteins. In this thesis, I present a new, and improved, formulation for ME-models, and apply it to elucidate the emergence of non-trivial elements of cell physiology. This new ME-model formulation, ETFL, allows the integration of more experimental data than the previous state of the art, while being more efficient than previously published equivalent methods. Then, I show ETFL elucidates complex cellular behaviors. In particular, I demonstrate the preferred consumption of specific carbohydrate by E. coli, or diauxie, is the result of an optimal program of the cell towards growth, under the constraints of proteome limitation. I show that ETFL can be adapted to elucidate the dynamics of the proteome in the cell and the transition from one physiology to another. I also describe the construction of a ME-model of a eukaryotic organism, S. cerevisiae, and how the model produced can account for the emergence of overflow metabolism, or the Crabtree effect. Finally, I build a model of human colon cancer, and present a formulation for ETFL that allows to account for regulatory interactions in ME-models. I use the model to reproduce the known mechanisms of action of the drug metformin, and show it has a dual, dose-dependent action. I also show how such models can be used to predict potential mechanisms of resistance against treatment. In a second part of this thesis, I present open source software pieces I developed and contributed to, to promote open science. This work outlines the potential for ME-models in systems biology, and shows how to use them to elucidate complex cellular physiologies. The methods presented in this work also show how these new and improved ME-models constitute a major step towards systematic, integrated whole-cell modeling.