Abstract 5679: A novel approach for antibody affinity maturation using AI-driven mutation prediction

Abstract Background: Therapeutic antibodies, characterized by their exceptional specificity and efficacy, have been a cornerstone of biotherapeutics development since their discovery. However, achieving high-affinity functional antibodies remains a significant challenge. Traditional affinity maturation methods, such as saturation mutagenesis combined with phage display screening, require extensive library construction and screening, which are time-consuming, costly, and present a high barrier to implementation. Here, we introduce a novel affinity screening approach leveraging AI-based deep learning networks to predict potential mutation sites that enhance affinity. Coupled with high-throughput expression systems, this approach effectively reduces screening scale and rapidly generates higher-affinity antibodies. Methods: We developed and trained a novel AI model based on deep learning networks to facilitate antibody affinity optimization. The model simulates antigen-antibody docking and predicts affinity improvements by mutating the parent antibody's CDRs, identifying hotspot residues. We constructed the hotspot mutations and high-throughput expression in 96-well deep plates. The expressed antibody in supernatants were captured by protein A, and screened using sandwich ELISA. Mutations showing at least a 1.2-fold increase in OD450 are selected in preliminarily screening for AI to design combination mutations. Results: For IgG antibodies, computational simulations reduced the initial single-point mutation library size by an order of magnitude, from 103 to 102 mutations, significantly decreasing library construction and screening costs. This approach enabled the completion of the initial screening within three weeks, with 10% of positive mutations exhibiting an OD450 increase of at least 1.2-fold. For VHH antibodies, optimization resulted in blocking activity increase up to a twofold reduction in IC50 and an 18.89-fold increase in affinity as measured by SPR. Similarly, IgG antibodies exhibited a 6.01-fold improvement in affinity (SPR). Conclusions: We developed an AI model to simulate antigen-antibody binding and identify mutations for affinity enhancement. These predicted mutations were subsequently validated through experimental screening, confirming improved affinity. This approach enables the rapid and cost-effective generation of high-affinity antibodies. Citation Format: Yiran Li, Hao Peng, Hui Zhao, Xiaoyi Han, Panpan Zhang, Jinying Ning, Feng Hao. A novel approach for antibody affinity maturation using AI-driven mutation prediction [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5679.