A Three-Long Non-Coding RNA Signature Derived from the Cancer Genome Atlas Database Predicts the Survival of Patients with Head and Neck Squamous Cell Carcinoma

Background: Long non-coding RNAs (lncRNAs) have important biological functions and can be used as prognostic biomarkers in cancer. Objectives: To identify a prognostic signature of lncRNAs for head and neck squamous cell carcinoma (HNSCC). Methods: We analysed the RNA-seq data derived from The Cancer Genome Atlas database to identify a prognostic lncRNA signature model by using the orthogonal partial least square discrimination analysis (OPLS-DA) and 1.5-fold expression change criterion methods. The prognosis prediction model constructed on the lncRNA signatures and clinical parameters were evaluated by using five-fold cross validation method. Findings: 84 out of 3199 lncRNAs were significantly associated with survival of patients with HNSCC (log-rank test 22 < 0.01). By using OPLS-DA and 1.5-fold change selection criterion, five lncRNAs (KTN1-AS1, LINC00460, GUSBP11, LINC00923 and RP3-894A10.6) were further selected. The prediction power of each combination of the five lncRNAs was evaluated through receiver operating characteristic (ROC) curve and a three-lncRNA panel (KTN1-AS1, LINC00460 and RP3-894A10.6) has achieved the highest prognostic prediction power (area under the curve, 0.68; 95% CI 0.60–0.76; P < 0.0001) in the cohort. Patients were categorised into high- and low-risk groups based on their three-lncRNA profiles. Patients with high-risk score had worse overall survival than those with low risk scores in the cohort (log-rank test, P = 0.0003). Univariable and multivariable Cox regression analysis showed that the lncRNAs signature and tumour grade were independent prognostic factors for patients with HNSCC. Conclusion: Our findings showed that the three-lncRNA signature might be a novel biomarker for accurate prediction prognosis of patients with HNSCC.