With advancements in high-throughput technologies and open availability of bioassay data, computational methods to generate models, that zoom out from a single protein with a focused ligand set to a larger and more comprehensive description of compound-protein interactions and furthermore demonstrate subsequent translational validity in prospective experiments, are of prime importance. In this article, we discuss some of the new benefits and challenges of the emerging computational chemogenomics paradigm, particularly with respect to compound-protein interaction. Examples of experimentally validated computational predictions and recent trends in molecular feature extraction are presented. In addition, analyses of cross-family interactions are considered. We also discuss the expected role of computational chemogenomics in contributing to increasingly expansive network-level modeling and screening projects.