Background: Nasopharyngeal carcinoma (NPC) is a highly invasive and metastatic cancer, with diverse molecular characteristics and clinical outcomes. This study aims to dissect the molecular heterogeneity of NPC, followed by the construction of a microRNA (miRNA)-based prognostic model for prediction of distant metastasis.
Methods: We retrieved two NPC datasets: GSE32960 and GSE70970 as training and validation cohorts, respectively. Consensus clustering was employed for cluster discovery, and support vector machine was used to build a classifier. Finally, Cox regression analysis was applied to constructing a prognostic model for predicting risk of distant metastasis.
Results: Three NPC subtypes (immunogenic, classical and mesenchymal) were identified that are molecularly distinct and clinically relevant, of which mesenchymal subtype (~ 36%) is associated with poor prognosis, characterized by suppressing tumor suppressor miRNAs and the activation of epithelial--mesenchymal transition. Out of the 25 most differentially expressed miRNAs in mesenchymal subtype, miR-142, miR-26a, miR-141 and let-7i have significant prognostic power (P < 0.05).
Conclusions: We proposed for the first time that NPC can be stratified into three subtypes. Using a panel of 4 miRNAs, we established a prognostic model that can robustly stratify NPC patients into high- and low- risk groups of distant metastasis.
Keywords: Consensus clustering; Cox regression model; Distant metastasis; Molecular subtyping; Nasopharyngeal carcinoma; microRNA.