Optimizing microbial hosts for large-scale production of valuable metabolites has two main challenges: (i) maximizing the expression and function of the proteins needed to perform the desired reactions; and, (ii) adapting host metabolism to support these new reactions and remove unwanted or toxic side products. Although considerable effort has been focused on the first challenge, the methodology developed and described in this dissertation addresses the second challenge. As a test case, I sought to increase production of L-DOPA, a pharmaceutically-relevant metabolite and precursor to the benzylisoquinoline (BIA) class of medicinal molecules, in the yeast S. cerevisiae. Production of L-DOPA and derived BIAs in yeast can be accomplished through the action of CYP76AD1, a heterologously-expressed cytochrome P450 enzyme from beet. Hence, I sought to devise strategies for isolating strain variants that carry genome modifications that improve CYP76AD1-dependent L-DOPA production. My approach was based on the assumption that endogenous factors in yeast restrain or impede this process. To perform multiple rounds of mutagenesis and screening, we constructed an in vitro barcoded transposon-disruption library. This library was introduced and integrated into the host genome by homologous recombination. The resulting variants were screened using a biosensor in which L-DOPA produced by the cells is converted to a fluorescent derivative via the action of the enzyme DOPA dioxygenase (DOD). Thus, I was able to use high-throughput fluorescence-activated cell sorting (FACS) to enrich for the desired variants. I conducted this approach, iteratively, for three rounds, i.e. improved strains obtained from the preceding rounds were transformed with the transposon-disruption library and re-screened. In the first two rounds of screening, I identified deletions that improved biosensor compartmentalization and, consequently, improved the reliability of the read-out for L-DOPA production. In the final round, I discovered that deletion of the gene encoding a heme oxygenase (HMX1) that is localized to the endoplasmic reticulum and involved in heme degradation increased both total cellular heme content and L-DOPA production (as monitored by measuring its derivative dopamine as a proxy). I demonstrated further that deleting HMX1 may represent a general strategy for improving the performance of heterologous P450 enzymes in yeast because the absence of Hmx1 also enhanced the ability of a second P450 enzyme, BM3, to generate its product.