Generative Humanization for Therapeutic Antibodies

  • 2024-12-09 02:26:45
  • Cade Gordon, Aniruddh Raghu, Peyton Greenside, Hunter Elliott
  • 0

Abstract

Antibody therapies have been employed to address some of today's mostchallenging diseases, but must meet many criteria during drug developmentbefore reaching a patient. Humanization is a sequence optimization strategythat addresses one critical risk called immunogenicity - a patient's immuneresponse to the drug - by making an antibody more "human-like" in the absenceof a predictive lab-based test for immunogenicity. However, existinghumanization strategies generally yield very few humanized candidates, whichmay have degraded biophysical properties or decreased drug efficacy. Here, were-frame humanization as a conditional generative modeling task, wherehumanizing mutations are sampled from a language model trained on humanantibody data. We describe a sampling process that incorporates models oftherapeutic attributes, such as antigen binding affinity, to obtain candidatesequences that have both reduced immunogenicity risk and maintained or improvedtherapeutic properties, allowing this algorithm to be readily embedded into aniterative antibody optimization campaign. We demonstrate in silico and in labvalidation that in real therapeutic programs our generative humanization methodproduces diverse sets of antibodies that are both (1) highly-human and (2) havefavorable therapeutic properties, such as improved binding to target antigens.

 

Quick Read (beta)

loading the full paper ...