AbNatiV: VQ-VAE-based assessment of antibody and nanobody nativeness for engineering, selection, and computational design

Monoclonal antibodies have emerged as a key class of therapeutics, and nanobodies are rapidly increasing in popularity following the approval of the first nanobody drug in 2019, yet the therapeutic development of these biologics remains a challenge. Computational design is a promising technology to accelerate antibody discovery and optimisation and to address some critical limitations. However, despite the availability of established in vitro directed evolution technologies that are relatively fast and cheap to deploy, the gold standard for generating therapeutic antibodies remains discovery from animal immunization or patients. Immune-system derived antibodies tend to have favourable properties in vivo, including long half-life, low cross reactivity with self-antigens, and low toxicity, which raises the question of whether computational design will ever be able to generate such antibodies. Here, we present AbNatiV, a deep-learning tool for assessing the nativeness of antibodies and nanobodies, i.e., their likelihood of belonging to the distribution of immune-system derived human antibodies or camelid nanobodies. AbNatiV is trained on curated datasets of human antibodies or camelid nanobodies and can accurately predict the nativeness of a given sequence. We show that AbNatiV can be used to design antibodies and nanobodies that are indistinguishable from immune-system derived ones, to facilitate humanization, and to predict the likelihood of immunogenicity. AbNatiV is a multi-purpose tool that will be valuable for developing antibodies and nanobodies with improved efficacy and safety profiles. We make it readily accessible to users through both downloadable software and as a webserver. ### Competing Interest Statement The authors have declared no competing interest.