Epistasis occurs when the phenotypic consequence of one mutation depends on the presence or absence of another mutation. Given that epistasis frequently uncovers functional association between genes, the global mapping of such interactions in model experimental systems could help understand the genetics underlying complex inherited phenotypes. To this end, we have developed a quantitative experimental approach in which we measure fitness of single and double mutants by direct competition of Saccharomyces cerevisiae strains tagged with cyan or yellow variants of the green fluorescent protein. Following this strategy in parallel experiments in a robotic system, we are building a comprehensive network of epistatic interactions for yeast metabolism. Interestingly, the quantitative nature of our approach allows the accurate scoring of negative -aggravating- and positive -buffering- epistasis. It has been recently pointed out that comprehensive empirical data for both negative and positive interactions would allow inferring functional gene modules from pure phenotypic measurements based on the exclusiveness of the inter-module interaction signs. By highlighting epistatic interactions between, rather than within, functional modules, our system-level approach extends the concept of genetic interaction from single genes to functional units.