Search

Antitumor peptide vaccine is an alternative and attractive way to eradicate tumors by the aid of host immune system. The current peptide vaccine development is hampered by several hurdles. These include heterogeneity of tumor cells, applicable in vivo stability of peptides, and others. One of the major obstacles is the precise prediction of cancer antigen, which can be loaded efficiently onto antigen presenting cells. We present here a neoantigen prediction algorithm DeepNeo comprised of DeepNeo-MHC and DeepNeo-TCR, which calculates the antigenicity of neoantigen based on the MHC binding and TCR activation potential. We examined the efficacy of DeepNeo using ELISPOT and two allogenic mouse tumor models. By treating the combination of long peptides targeting MC38 colon or B16F10 melanoma cell lines, we have shown the DeepNeo effectively determines neoantigens for the development of antitumor peptide vaccine. Analysis of tumor-infiltrated lymphocytes (TIL) supported our results. Undergoing study using patient-derived xenograft model treated with autologous, trained PBMC with personalized peptide vaccine will be discussed further. Citation Format: Eun Ji Lee, Min Ji Park, Soo Jin Kim, Seung-Jae Noh, Inkyung Shin, Jeong Yeon Kim, Incheol Shin, Jung Kyoon Choi, Dae-Yeon Cho, Suhwan Chang. Efficacy analysis of the DeepNeo neoantigen prediction tool for antitumor vaccine development [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 2 (Clinical Trials and Late-Breaking Research); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(8_Suppl):Abstract nr LB168.

Previous biomarkers such as PD-L1 expression, tumor mutation burden (TMB), and neoantigen load have limited accuracy in predicting responses to immune checkpoint inhibitor (ICI) therapy. In this work, we propose an improved predictor of ICI responses on the basis of our previous findings on the role of neoantigen functionality and genomic hypomethylation in tumor immunity. Specifically, we developed a metric of constitutive neoantigen load given the association of constitutive neoantigens derived from genes indispensable for cancer growth with favorable clinical responses to ICI (Clin. Transl. Med. 12:e714). A panel that captures somatic mutations on the 300 most essential genes identified by genome-wide screening in ~1,800 pan-cancer cell lines was used. Our DeepNeo algorithm (Nat. Commun. 11:951; Nat. Genet. 2023) was employed to identify T cell-reactive neoantigens from somatic mutations binding patient-matched HLA proteins. Also, considering the role of genomic hypomethylation in inducing resistance of tumors to ICI therapy (Nat. Commun. 10:4278), we developed an assay that estimates genomic hypomethylation by targeted sequencing of LINE-1 elements. Constitutive neoantigen load and LINE-1 methylation status were evaluated in our cohort consisting of 335 lung cancer patients treated with ICIs. As a result, constitutive neoantigen load (HR=0.671; p-value=3.13E-03) and LINE-1 methylation (HR=0.597; p-value=0.012) outperformed PD-L1 (HR=0.74; p-value=0.199), TMB (HR=0.917; p-value=0.511), and the combination of PD-L1 and TMB (HR=0.707; p-value=0.218). The combination of constitutive neoantigen load and LINE-1 methylation (HR=0.245; p-value=1.11E-05) showed even better predictive power. In conclusion, our approach can be used as new companion diagnostics for ICI-based immune-oncology. Citation Format: Hyoeun Bang, Jeong Yeon Kim, Cheolyong Joe, Kyeonghui Kim, Younghak Bang, Hyemin Kim, Honghi Cha, Eunjoo Oh, Naeun Lee, Inkyung Shin, Seung-Jae Noh, Dae-Yeon Cho, Se-Hoon Lee, Jung Kyoon Choi. Improved prediction of immunotherapy responses by constitutive neoantigen load and LINE-1 methylation. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5475.