A Large Language Model Guides the Affinity Maturation of Variant Antibodies Generated by Combinatorial Optimization

Abstract

The ability of an antibody to bind an antigen with high specificity and strength (i.e., its binding affinity) are critical properties in the design of neutralizing antibodies. Recent technical advances in AI and a surge of experimental data on antigen-antibody interaction are driving innovations in the design and optimization of antibodies via affinity maturation. Here we introduce Ab-Affinity, a novel large language model which can accurately predict the binding affinity of specific antibodies against a target peptide within the SARS-CoV-2 spike protein. When used in conjunction with a genetic algorithm and simulated annealing, Ab-Affinity can generate novel antibodies with more than a 160-fold increase in predicted binding affinity compared to those obtained experimentally. Our experimental results show that the synthetic antibodies produced by Ab-Affinity have strong predicted biophysical properties. Molecular docking and molecular dynamics simulation of binding interactions of the best synthetic antibodies show enhanced interactions and stability on the target peptide epitope. In general, antibodies generated by Ab-Affinity are superior to those obtained with other existing computational methods.