Abstract 3538: The development of neoantigen-derived peptide vaccine driven by AI epitope design and computer assisted epitope enhancement

Abstract Background and objective: Neoantigens are the ideal targets for developing cancer vaccine for cancer treatment. However, the development of cancer vaccines is often hindered by poor antigen presentation of the vaccine peptide by major histocompatibility complex (MHC) molecules. Our AI epitope prediction model showed prediction accuracy superior to several state-of-the-art algorithms for MHC-I antigen presentation. As studies showed that amino acid substitution in key positions of MHC-I epitope provides strong activating signals to cytotoxic T cells, we aimed to develop MHC-I restricted neoantigen vaccines by coupling our AI model with computer-assisted modification of neoantigen-derived epitopes to improve their HLA presentation probability and immunogenicity. Methods: The AI model was used to analyze some common TP53 neoantigen sequences and predict epitopes for HLA-A*02:01. The epitopes underwent computer-assisted saturation mutagenesis that simulated single amino acid substitutions to obtain the associated presentation score changes. Peptide sequences with improved presentation scores were synthesized, and their binding affinities with HLA-A*02:01 were analyzed by surface plasmon resonance (SPR). The peptides were immunized into the HLA-A*02:01-humanized C57BL/6 mice, and their immunogenicity was determined by mouse interferon γ (mIFNγ) ELISpot assay and intracellular cytokine staining (ICS) using splenocytes of the mice. Results: Our AI model predicted epitopes from common TP53 neoantigens that showed high HLA-A*02:01 presentation probability. Through saturation mutagenesis, dominant mutations with at least 5% increase of presentation probability were screened out. Seven novel modified peptides were synthesized and showed high binding affinity with HLA-A*02:01/ß2-microglobulin heterodimer. After these peptides were pooled to immunize HLA-A*02:01-humanized mice, mIFNγ ELISpot assay showed that 6 out of the 7 modified peptides reactivated splenocytes from the peptide-immunized mice, while mIFNγ ICS showed that CD8+ cytotoxic T-cells were activated. The splenocytes were then restimulated with the native neoantigen peptides corresponding to the modified peptides. One native neoantigen peptide reactivated T-cells from the immunized mice, suggesting that immunization of the modified peptides could induce cytotoxic T-cell response against the native neoantigens. Conclusion: Our study demonstrates a highly efficient approach by coupling MHC-I presentation prediction AI model and computer assisted saturation mutagenesis to design single amino acid substituted MHC-I peptides derived from neoantigens. Immunization of the peptides induced cytotoxic T-cell responses which may inhibit the growth of cancer cells carrying such a neoantigen. This novel approach to developing cancer vaccines will be further tested in future studies. Citation Format: Chi Han Samson Li, Hong Wang, Kin Tak Chan, Genwei Zhang, Jiewen Du, Lipeng Lai, Melvin Toh. The development of neoantigen-derived peptide vaccine driven by AI epitope design and computer assisted epitope enhancement [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 3538.