Identification of a nomogram based on an 8-lncRNA signature as a novel diagnostic biomarker for head and neck squamous cell carcinoma

Abstract

Long noncoding RNAs (lncRNAs) have been proposed as diagnostic or prognostic biomarkers of head and neck squamous carcinoma (HNSCC). The current study aimed to develop a lncRNA-based prognostic nomogram for HNSCC. LncRNA expression profiles were downloaded from The Cancer Genome Atlas (TCGA) database. After the reannotation of lncRNAs, the differential analysis identified 253 significantly differentially expressed lncRNAs in training set TCGA-HNSC (n = 300). The prognostic value of each lncRNA was first estimated in univariate Cox analysis, and 41 lncRNAs with P < 0.05 were selected as seed lncRNAs for Cox LASSO regression, which identified 11 lncRNAs. Multivariate Cox analysis was used to establish an 8-lncRNA signature with prognostic value. Patients in the high-signature score group exhibited a significantly worse overall survival (OS) than those in the low-signature score group, and the area under the receiver operating characteristic (ROC) curve for 3-year survival was 0.74. Multivariable Cox regression analysis among the clinical characteristics and signature scores suggested that the signature is an independent prognostic factor. The internal validation cohort, external validation cohort, and 102 HNSCC specimens quantified by qRT-PCR successfully validate the robustness of our nomogram.

Keywords: bioinformatics; head and neck squamous carcinoma; long noncoding RNAs; prognosis; qRT-PCR.

Figure 1

Volcano plot of the differentially expressed mRNAs and lncRNAs between HNSCC and para-carcinoma tissues. Red indicates high expression, and blue indicates low expression (|log2FC| > 1 and P value < 0.05). The Y axis represents adjusted P values, and the X axis represents log2FC values. The RNAs studied in this article have been marked in the figure. (A) Volcano plot of the differentially expressed lncRNAs. (B) Volcano plot of the differentially expressed mRNAs.

Figure 2

Establishment and validation of the eight-lncRNA prognostic signature. (A–C) The procedure of establishing the prognostic signature. (D) Correlation between the prognostic signature and the overall survival of patients in the TCGA cohort. The distribution of signature scores (top), survival time (middle) and lncRNA expression levels (bottom). The black dotted lines represent the median signature score cut-off dividing patients into the low- and high-signature groups. The red dots and lines represent the patients in the high-score group. The green dots and lines represent the patients in the low-score group. (E) Kaplan-Meier curves of OS based on the 8-lncRNA signature. (F) ROC curve analyses based on the 8-lncRNA signature.

Figure 3

Screening of prognosis-related clinical characteristics by Kaplan-Meier analysis. (A) Kaplan-Meier curves based on different age groups, where Q1, Q2, Q3, and Q4 represent quartiles. (B) Kaplan-Meier curves based on gender. (C) Kaplan-Meier curves based on different N stages. (D) Kaplan-Meier curves based on different M stages. (E) Kaplan-Meier curves based on new events. (F) Kaplan-Meier curves based on different tumor stages.

Figure 4

Construction of a nomogram for overall survival prediction in HNSCC. (A) Univariate and multivariate Cox regression analyses of clinical factors associated with overall survival. (B) The nomogram consists of M stage, new event, stage and the signature score based on the eight-lncRNA signature. (C) ROC curves according to the nomogram and lncRNA signature score. (D) Calibration curves of the nomogram for the estimation of survival rates at 3 and 5 years. (E) Kaplan-Meier curves of OS according to the total risk score.

Figure 5

Validation of the model by the internal validation set TCGA-HNSCC (n=199). (A) Distribution of 8-lncRNA-based signature scores, lncRNA expression levels and patient survival durations in the internal validation set. (B) Kaplan-Meier curves of OS based on the 8-lncRNA signature. (C) ROC curve analyses based on the 8-lncRNA signature. (D) ROC curves according to the nomogram and lncRNA signature score. (E) Calibration curves of the nomogram for the estimation of survival rates at 3 and 5 years. (F) Kaplan-Meier curves of OS according to the total risk score.

Figure 6

Validation of the model by the external validation set GSE65858 (n=270). (A) Distribution of 8-lncRNA-based signature scores, lncRNA expression levels and patient survival durations in the external validation set. (B) Kaplan-Meier curves of OS based on the 8-lncRNA signature. (C) ROC curve analyses based on the 8-lncRNA signature. (D) ROC curves according to the nomogram and lncRNA signature score. (E) Calibration curves of the nomogram for the estimation of survival rates at 3 and 5 years. (F) Kaplan-Meier curves of OS according to the total risk score.

Figure 7

Validation of the model by the qRT-PCR set (n=102). (A) Distribution of 8-lncRNA-based signature scores, lncRNA expression levels and patient survival durations in the qRT-PCR validation set. (B) Kaplan-Meier curves of OS based on the 8-lncRNA signature. (C) ROC curve analyses based on the 8-lncRNA signature. (D) Calibration curves of the nomogram for the estimation of survival rates at 2 and 3 years. (E) Kaplan-Meier curves of OS according to the total risk score. (F) ROC curves according to the nomogram and lncRNA signature score.

Figure 8

WGCNA. (A) Analysis of the scale-free topology model fit index for various soft-thresholding powers (β) and the mean connectivity for various soft-thresholding powers. Overall, 3 was the most fitting power value. (B) Dendrogram of the genes and different clinical factors of HNSCC (survival time, survival status, sex, age, grade, stage, T stage, N stage, M stage, new event, signature score). (C) Dendrogram of the gene modules based on a dissimilarity measure. The branches of the cluster dendrogram correspond to the different gene modules. Each piece of the leaves on the cluster dendrogram corresponds to a gene. (D). Module-trait relationships. Heatmap of the correlation between module eigengenes and clinical characteristics of HNSCC. (E) Hierarchical clustering and heatmap of the hub gene network.

Figure 9

The correlation between the genes in the modules and survival time. (A) Distribution of mean gene significance and standard deviation with survival time in the HNSCC modules. (B) Scatter plot of module eigengenes in red and brown modules. GO (C–E) and KEGG (F) pathway enrichment of eight modules. GO enrichment contains three categories: biological process (C), cellular component (D) and molecular function (E).

Figure 10

The correlation between the genes in the modules and grade. (A) Distribution of mean gene significance and standard deviation with grade in the HNSCC modules. (B) Scatter plot of the module eigengenes in the turquoise, black, and red modules. (C) The lncRNA-mRNA network (weight>0.1) of the hub lncRNAs in the turquoise module. Red and blue diamond shapes represent up- and downregulated lncRNAs, respectively. Purple circles represent mRNAs. GO (D–F) and KEGG (G) pathway enrichment of eight modules. GO enrichment contains three categories: biological process (D), cellular component (E) and molecular function (F).