Physical interactions between proteins are essential for biological processes. Hence, there have been major efforts to elucidate the complete networks of protein-protein interactions, or "interactomes", of various organisms. Detailed descriptions of protein interaction networks should include information on the discrete domains that mediate these interactions, yet most large-scale efforts model interactions between whole proteins only. We previously developed a yeast two-hybrid-based strategy to systematically map interaction domains and generated a domain-based interactome network for 750 proteins involved in C. elegans early embryonic development. Here, we expand the concept of Y2H-based interaction domain mapping to the genome-wide level. We generated a human fragment library by randomly fragmenting the full-length open reading frames (ORFs) present in the human ORFeome collection. Screens using several proteins required for cell division or polarity establishment as baits demonstrate the ability to accurately identify interaction domains for human proteins using this approach, while the experimental quality of the Y2H data was independently verified in coaffinity purification assays. The library generation strategy can easily be adapted to generate libraries from full-length ORF collections of other organisms.