Efficient evolution of human antibodies from general protein language models

This study demonstrates that general protein language models can efficiently evolve human antibodies by suggesting mutations that are evolutionarily plausible, even without information about the target antigen, binding specificity, or protein structure. Affinity maturation of seven antibodies was performed, screening 20 or fewer variants per antibody across two rounds of laboratory evolution. Binding affinities were improved up to sevenfold for four clinically relevant antibodies and up to 160-fold for three unmatured antibodies. Many designs showed favorable thermostability and viral neutralization activity. The models' effectiveness extended to diverse protein families and selection pressures, suggesting broad applicability.

Brian L. Hie, Varun R. Shanker, Duo Xu, Theodora U. J. Bruun, Payton A. Weidenbacher, Shaogeng Tang, Wesley Wu, John E. Pak, Peter S. Kim