BPS2026 – An iterative de novo antibody design pipeline for affinity optimization against Coxsackievirus A6 capsid proteins

Antibodies have a characteristic structure with a set of complementarity-determining regions whose sequences govern the binding affinity of an antibody for a particular epitope. Recent advances have enabled the design of proteins using machine-learning-based algorithms whose computational designs demonstrate remarkable efficacy in vitro and in vivo. Despite this progress, it remains unclear how to best harness these tools for understanding the biochemical mechanisms of effective therapeutic proteins. Here, we developed a computational framework for antibody design, which generates novel candidates with progressively improved affinity for Coxsackievirus A6 capsid proteins over multiple iterations. We found that a per-residue level energetics-informed design process is critical to generating antibodies with high affinity for their respective epitopes and investigated the key residue-residue interactions stabilizing the antibody-capsid complex. Augmentations to machine-learning-based algorithms such as the energy-based analysis shown here are expected to be integral to the future of computational protein therapeutic design.