Document resource
Background Cancer immunotherapy treatments, like cancer vaccines, yield tremendous results in tumor regression for many patients. However, Black and Latine patients can experience increased rates of treatment failure and toxicities. 1–4 Biases seen in Human Leukocyte Antigen (HLA) binding algorithms (e.g., NetMHCPan4.1) may contribute to their experienced treatment failure.5 6 Additionally, neoantigens are patient specific and often exhibit low immunogenicity.7 Alternatively, researchers demonstrated that altered peptide ligands (APLs), synthesized from original neoantigens, can increase antitumor activity against the original neoantigens presented on tumoral cells.8 9 Therefore, our goal is to generate safe, efficacious altered peptide ligands (APLs), from gathered PIK3CA-associated neoantigens, to create generalized cancer vaccine candidates for Black and Latine patients. We hypothesized that structure-based methods can complement sequence-based algorithms in designing APLs with improved HLA-binding, higher immunogenicity, and lower off-target toxicity risks.Methods The Allele Frequency Net Database 10 enabled the identification of prevalent HLA-I alleles for Black and Latine patients in the United States. PIK3CA-associated neoantigens were sourced from the Immune Epitope Database11 and Tumor-Specific Neoantigen Database2.0.12 HLA-binding predictions for neoantigen-HLA complexes were done via BigMHC,13 NetMHCPan4.1,14 and MHCFlurry2.0.15 Each complex was modeled with APE-Gen2.0.16 The stability, immunogenicity, and off-target toxicity for each neoantigen-HLA was predicted with TL-MHC17 and CrossDome.18 We developed PPmGen to generate APLs with single and double-point mutations for each neoantigen. Finally, we compared the binding, stability, immunogenicity, structural similarities, and off-target toxicity risks between the original neoantigen-HLAs and APL-HLAs.Results Few validated neoantigens were found for the HLAs of interest ( figure 1). Additionally, each HLA-binding algorithm selected false positives from the negative controls for all HLAs of interest, with MHCFlurry2.0 exhibiting a higher false positive rate. BigMHC and NetMHCPan4.1 were used in consensus to identify 16 putative neoantigen-HLAs. Each complex showed low efficacy. PPmGen produced over 20,000 APLs, with 4,057 APLs exhibiting high biochemical similarity, enabling the binding prediction and modeling of nearly 80,000 APL-HLAs. Of the 16 neoantigen-HLA complexes, we yielded 15 structurally and biochemically similar APL-HLAs with improved efficacy, with 6 overlapping APL-HLAs demonstrating improved efficacy and off-target toxicity (figure 2). We are currently experimentally validating our 6 APL-HLA candidates.Conclusions In this project, we created a novel immunoinformatics pipeline for APL design and identified high-quality APLs with improved efficacy and reduced off-target toxicity risk. This work may enhance and accelerate cancer vaccine design, regardless of patient demographics and cancer types.Acknowledgements Acknowledgments Research reported in this publication was supported by the National Institute on Minority Health and Health Disparities (NIMHD) of the National Institutes of Health (NIH) to the University of Houston under Award Number U54MD015946. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.References Chandani KU, Agrawal S, Raval M, Khan H. CN28 racial disparities in immune-related adverse events in patients with lung cancer treated with immune checkpoint inhibitors. Ann Oncol. 2024;35:S1176.Freeman JQ, Huo D, Shubeck SP, Chen N, Yarlagadda SR, Nanda R, Howard FM. Trends and disparities in the use of immunotherapy for triple-negative breast cancer in the US. JAMA Netw Open. 2025;8(2):e2460243.Klingbeil KD, Dillon DL, Zarrinkhoo E, Bechay K, Park JY, Rook JM, et al. Hispanic/Latino Ethnicity is an independent predictor of worse survival for gastric cancer in a multicenter safety-net patient population. Cancer Epidemiol Biomarkers Prev. 2025;34(1):75–84.Al-Sadawi M, Hussain Y, Copeland-Halperin RS, Tobin JN, Moskowitz CS, Dang CT, et al. Racial and socioeconomic disparities in cardiotoxicity among women with HER2-positive breast cancer. Am J of Cardiol. 2021;147:116–121.Norori N, Hu Q, Aellen FM, Faraci FD, Tzovara A. Addressing bias in big data and AI for health care: a call for open science. Patterns. 2021;2(10):100347.Dankwa-Mullan I, Weeraratne D. Artificial intelligence and machine learning technologies in cancer care: addressing disparities, bias, and data diversity. Cancer Discov. 2022 Jun 2;12(6):1423–1427.Xie N, Shen G, Gao W, Huang Z, Huang C, Fu L. Neoantigens: promising targets for cancer therapy. Signal Transduct Target Ther. 2023;8(1):9.Cole DK, Edwards ESJ, Wynn KK, Clement M, Miles JJ, Ladell K, et al. Modification of MHC anchor residues generates heteroclitic peptides that alter TCR binding and T cell recognition. J Immunol. 2010;185(4):2600–2610.Slansky JE, Nakayama M. Peptide mimotopes alter T cell function in cancer and autoimmunity. Semin in Immunol. 2020;47:101395.Gonzalez-Galarza FF, McCabe A, Santos EJMD, Jones J, Takeshita L, Ortega-Rivera ND, et al. Allele frequency net database (AFND) 2020 update: gold-standard data classification, open access genotype data and new query tools. Nucleic Acids Res. 2019;48(D1):D783-D788.Vita R, Blazeska N, Marrama D, IEDB Curation Team Members, Shackelford D, Zalman L, et al. The immune epitope database (IEDB): 2024 update. Nucleic Acids Res. 2025;53(D1):D436–443.Wu J, Chen W, Zhou Y, Chi Y, Hua X, Wu J, et al. TSNAdb v2.0: the updated version of tumor-Specific neoantigen database. Genomics Proteomics Bioinformatics. 2023;21(2):259–266.Albert BA, Yang Y, Shao XM, Singh D, Smith KN, Anagnostou V, et al. Deep neural networks predict class I major histocompatibility complex epitope presentation and transfer learn neoepitope immunogenicity. Nat Mach Intell. 2023;5(8):861–872.Reynisson B, Alvarez B, Paul S, Peters B, Nielsen M. NetMHCpan-4.1 and NetMHCIIpan-4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data. Nucleic Acids Res. 2020;48(W1):W449–454.O’Donnell TJ, Rubinsteyn A, Laserson U. MHCflurry 2.0: improved pan-allele prediction of MHC class I-presented peptides by incorporating antigen processing. Cell Syst. 2020;11(4):418–419.Fasoulis R, Rigo MM, Lizée G, Antunes DA, Kavraki LE. APE-Gen2.0: expanding rapid class I peptide-major histocompatibility complex modeling to post-translational modifications and noncanonical peptide geometries. J Chem Inf Model. 2024;64(5):1730–1750.Fasoulis R, Rigo MM, Antunes DA, Paliouras G, Kavraki LE. Transfer learning improves pMHC kinetic stability and immunogenicity predictions. ImmunoInformatics. 2024;13:100030.Fonseca AF, Antunes DA. CrossDome: an interactive R package to predict cross-reactivity risk using immunopeptidomics databases. Front Immunol. 2023;14:1142573.Abstract 1294 Figure 1Neoantigen distribution per HLA allele for PIK3CA-associated breast cancer samples. A. Only 12 validated neoantigens were found in IEDB for PIK3CA-associated breast and pan-cancer samples. B. Only 42 predicted neoantigens were found in TCGA dataAbstract 1294 Figure 2Comparing the efficacy of the APL candidates to the original neoantigens. A-C. Binding, immunogenicity, and stability predictions between the 16 original neoantigens and the best APL candidates were compared. D. Overlap of the best APL-HLAs