The biological function of proteins may be predicted by identification of their interacting partners, and one of the major goals of the postgenomic era is the mapping of protein interaction networks. Membrane proteins are of particular interest because of their role in disease and because of their prevalence as major pharmaceutical targets. Unfortunately, because of their hydrophobic nature, they have long been difficult to study in a high-throughput format. A powerful technology recently developed to facilitate the characterization of membrane protein interactions is the membrane yeast two-hybrid (MYTH) assay. MYTH adapts the principle of split ubiquitin for use as a potent in vivo sensor of protein-protein interactions, allowing large-scale screening for interactors of full-length membrane proteins, from a range of organisms, using Saccharomyces cerevisiae as a host. In this article, we describe a protocol for MYTH bait generation, validation and library screening. The entire MYTH procedure can generally be completed in 4-6 weeks.