We propose to elucidate the structural design principles of naturally occurring antibody complementarity-determining regions (CDRs) and to computationally design novel antibody functions. Antibodies represent the most versatile known system for molecular recognition. Research has yielded many insights into antibody design principles and promising biotechnological and pharmaceutical applications. Still, our understanding of how CDRs encode specific loop conformations lags far behind our understanding of structure-function relationships in non-immunological scaffolds. Thus, design of antibodies from first principles has not been demonstrated. We propose a computational-experimental strategy to address this challenge. We will: (a) characterize the design principles and sequence elements that rigidify antibody CDRs. Natural antibody loops will be subjected to computational modeling, crystallography, and a combined in vitro evolution and deep-sequencing approach to isolate sequence features that rigidify loop backbones; (b) develop a novel computational-design strategy, which uses the >1000 solved structures of antibodies deposited in structure databases to realistically model CDRs and design them to recognize proteins that have not been co-crystallized with antibodies. For example, we will design novel antibodies targeting insulin, for which clinically useful diagnostics are needed. By accessing much larger sequence/structure spaces than are available to natural immune-system repertoires and experimental methods, computational antibody design could produce higher-specificity and higher-affinity binders, even to challenging targets; and (c) develop new strategies to program conformational change in CDRs, generating, e.g., the first allosteric antibodies. These will allow targeting, in principle, of any molecule, potentially revolutionizing how antibodies are generated for research and medicine, providing new insights on the design principles of protein functional sites.