Toward real-world automated antibody design with combinatorial Bayesian optimization

Antibodies are multimeric proteins capable of highly specific molecular recognition. The complementarity determining region 3 of the antibody variable heavy chain (CDRH3) often dominates antigen-binding specificity. Hence, it is a priority to design optimal antigen-specific CDRH3 to develop therapeutic antibodies. The combi-natorial structure of CDRH3 sequences makes it impossible to query binding-affinity oracles exhaustively. Moreover, antibodies are expected to have high target specificity and developability. Here, we present AntBO, a combinatorial Bayesian optimization framework utilizing a CDRH3 trust region for an in silico design of anti-bodies with favorable developability scores. The in silico experiments on 159 antigens demonstrate that AntBO is a step toward practically viable in vitro antibody design. In under 200 calls to the oracle, AntBO suggests an-tibodies outperforming the best binding sequence from 6.9 million experimentally obtained CDRH3s. Addition-ally, AntBO finds very-high-affinity CDRH3 in only 38 protein designs while requiring no domain knowledge.

Automated summary

Antibodies are proteins capable of specific molecular recognition. The CDRH3 region of the antibody plays a crucial role in antigen-binding specificity. Designing optimal antigen-specific CDRH3 is important for therapeutic antibody development. In this study, we present AntBO, a Bayesian optimization framework that utilizes a CDRH3 trust region for in silico antibody design with favorable developability scores. Our results demonstrate that AntBO outperforms experimentally obtained CDRH3s and can identify high-affinity CDRH3 sequences with minimal protein designs and no domain knowledge.