Recombinant protein technologies have emerged as important tools for the production of proteins with industrial, academic and biopharmaceutical applications. However, current process development for a target protein is hindered by the burden recombinant protein expression places on the host system, with the level of this negative effect and the ideal production conditions varying from protein to protein. As a result current process optimisation relies on trial and error to determine the optimal set up for a given target protein. The work presented here will examine two mechanisms by which this burden acts on the cell, contributing towards making the overall burden effect and the optimal process conditions more predictable.Flux Balance Analysis was used to examine the effect of amino acid supplementation on the metabolic cost of a recombinant protein to predict which supplements would improve the efficiency of production, predictions supported elsewhere in the literature. However, experimental validation in batch and fed batch cultures for the production of human Granulocyte Colony Stimulating Factor demonstrated that these supplementation strategies do not lead to an increase in yield or performance. The results from the computational modelling alongside similar studies in the literature suggest that the more important factor may be optimisation for better growth generally rather than targeted attempts based on the protein composition.The sensitivity of native E. coli proteins to a loss of chaperone activity was predicted using the solubility data of the eSol database, identifying rrmJ as a protein of interest. The possible significance of rrmJ for chaperone saturation was examined alongside examining the effect of recombinant protein solubility using Green Fluorescent Protein (GFP) and its mutant GFP_A which have differing solubility. However, neither an effect through rrmJ nor a negative effect of recombinant protein solubility on growth was identified. Kinetic modelling for a mechanistic examination of the chaperone network suggests this is because the poorly folding protein is preferentially shifted to the insoluble fraction while the better performing proteins, i.e. the native proteins, are relatively unperturbed despite saturation of the chaperones.Overall the study was not able to make these areas more predictable. However, the observations made within this study contribute to an improvement in our understanding of two key mechanisms of interest in the field. Of particular interest is the identification of two logical hypotheses within the literature to be, at least in the cases tested, false.