A novel risk score model incorporating six co-stimulatory molecules for accurate prognosis prediction of laryngeal cancer.

Abstract

Background: Laryngeal cancer (LC) is a common respiratory tract malignancy. Although early-stage LC often responds well to treatment, advanced cases typically have poor outcomes and prognosis, resulting in a low overall survival (OS) rate. This study aimed to explore the correlation between co-stimulatory molecules and immune infiltration in LC and to construct a risk score (RS) model for predicting patient prognosis.

Methods: The RNA sequencing (RNA-seq) data of LC samples were downloaded from The Cancer Genome Atlas (TCGA) and used as the training dataset. The GSE27020 dataset served as the validation dataset. Univariate Cox regression analysis was performed to identify immune-related co-stimulatory molecules, based on which the samples were classified into three subtypes. Kaplan-Meier (KM) survival analysis was conducted to predict the survival prognosis in different subtypes. A prognostic RS model was constructed using the co-stimulatory molecules, which were obtained from the least absolute shrinkage and selection operator (LASSO) algorithm and validated using the GSE27020 dataset.

Results: Eighteen immune-co-stimulatory molecules were identified, allowing classification of the samples into three subtypes, among which subtype 2 exhibited the most favorable prognosis. Eight immune cell types were found to be associated with the subtypes, and ten immune checkpoint genes showed differential expression across them. Six optimized co-stimulatory molecules were selected to construct the RS model, which was capable of predicting LC prognosis with an area under the curve (AUC) value of 0.870 for 1-year survival in the TCGA dataset. Validation using GSE27020 yielded an AUC of 0.736.

Conclusions: An RS model incorporating six optimized co-stimulatory molecules was constructed and validated, demonstrating strong predictive power for the prognosis of patients with LC.