Computational design of structurally diverse antibody repertoires with drug-like properties

Animal immunization is the prevalent strategy for discovering antibody therapeutics1, but it is a lengthy and poorly controlled process. As an alternative, synthetic antibody repertoires deliver antibodies without animal welfare concerns, but the resulting antibodies often fail to exhibit "drug-like" biophysical properties1,2. Modern repertoires have improved developability by using a handful of frameworks with desirable biophysical properties, but at the cost of reduced structural diversity3-6. We developed a principled structure- and energy-based strategy, called CADAbRe, to navigate the complex tradeoffs between developability and structural diversity in repertoire design. The designed repertoire comprises billions of antibodies that are predicted to be stable and foldable, built from hundreds of different frameworks and hundreds of thousands of designed CDR H3s. We also developed an economical and scalable strategy for synthesizing large antibody repertoires, and as a proof of concept, designed and synthesized a 500-million variant phage display repertoire. Selections against four unrelated targets produced structurally diverse binders that exhibited drug-like properties, two of which were readily formatted as bispecifics for functional studies. Furthermore, one of the binders targets a challenging, highly charged surface. The proof-of-concept repertoire is available for academic research. We envision that the CADAbRe approach and repertoire will accelerate and rationalize antibody discovery while addressing animal-welfare concerns7.