The incidence of thyroid cancer (TC) increases year by year. It is necessary to construct a prognostic model for risk stratification and management of TC patients. Glutamine metabolism is essential for tumor progression and the tumor microenvironment. The present study aimed to develop a predictive model for TC using a glutamine metabolism gene set. Differentially expressed genes in cells with high glutamine metabolism levels from single cell RNA‑sequencing data were compared with genes differentially expressed between normal and TC tissues from The Cancer Genome Atlas Program data. Through Boruta feature selection methods and multivariate Cox regression, six crucial genes were identified for a risk‑scoring system to develop a prognostic model. The role of each gene was verified in TC cells in vitro. A risk‑scoring system was developed according to the glutamine gene set to forecast the overall survival of TC patients. This risk score could stratify TC patients and minimize unnecessary surgeries and invasive treatments. In addition, signal induced proliferation associated 1 like 2 (SIPA1L2), an important gene in the prognostic model, knockdown in TPC‑1 and BCPAP cell lines enhanced TC cell proliferation, migration and invasion. A risk model was developed based on a glutamine metabolism gene set. The model has reference values for TC stratification.
Keywords: Boruta; RNA‑seq; glutamine metabolism; multivariate Cox regression; thyroid cancer.