Background: The incidence of thyroid carcinoma (THCA), the most common endocrine tumor, is continuously increasing worldwide. Although the overall prognosis of THCA is good, patients with distant metastases exhibit a mortality rate of 5-20%.
Methods: To improve the diagnosis and overall prognosis of patients with thyroid cancer, we screened specific candidate neoantigen genes in early- and late-stage THCA by analyzing the transcriptome and somatic cell mutations in this study.
Results: The top five early-stage neoantigen-related genes (NRGs) were G protein-coupled receptor 4 [GPR4], chondroitin sulfate proteoglycan 4 [CSPG4], teneurin transmembrane protein 1 [TENM1], protein S 1 [PROS1], and thymidine kinase 1 [TK1], whereas the top five late-stage NRGs were cadherin 6 [CDH6], semaphorin 6B [SEMA6B], dysferlin [DYSF], xenotropic and polytropic retrovirus receptor 1 [XPR1], and ABR activator of RhoGEF and GTPase [ABR]. Subsequently, we used machine learning models to verify their ability to screen NRGs and analyze the correlations among NRGs, immune cell types, and immune checkpoint regulators. The use of candidate antigen genes resulted in a better diagnostic model (the area under the curve [AUC] value of the early-stage group [0.979] was higher than that of the late-stage group [0.959]). Then, a prognostic model was constructed to predict NRG survival, and the 1-, 3- and 5-year AUC values were 0.83, 0.87, and 0.86, respectively, which were closely related to different immune cell types. Comparison of the expression trends and mutation frequencies of NRGs in multiple tumors revealed their potential for the development of broad spectrum therapeutic drugs.
Conclusion: In conclusion, the candidate NRGs identified in this study could potentially be used as therapeutic targets and diagnostic biomarkers for the development of novel broad spectrum therapeutic agents.
Keywords: immune; machine learning; neoantigens; prognosis; thyroid carcinoma.
Copyright © 2023 Jia, Liang, Li, Qin, Li, Wang and Lu.