Identification of NRlncRNAs
The depiction of the intricate steps involved in the risk model and follow-up investigations is presented in Fig. 1. Specifically, the use of co-expression analysis recognized a total of 705 NRlncRNAs, and the network analysis of co-expressed genes related to necroptosis events is shown in Fig. 2A. Furthermore, following a comprehensive screening process, 12 lncRNAs were identified as key contributors to the diagnostic model, as outlined in Fig. 2B.
Construction and validation of signature
We evaluated the potential clinical utility of a prognostic model developed using the expression profiles of ncRNAs. We identified a total of 705 NRlncRNAs and used univariate Cox analysis to identify 74 prognostic NRlncRNAs (Fig. S1). The LASSO analysis was then applied for variable selection, and 24 NRlncRNAs were selected from the 74 prognostic NRlncRNAs (Fig. 3A and 3B). To develop a prognostic risk model, a multivariate Cox analysis was conducted with 12 NRlncRNAs (Fig. 3C). Survival analysis is an essential tool in clinical research to investigate the impact of different risk factors on patient outcomes. Our analysis of survival rates displayed higher SP levels in the low-risk cluster, signifying the effectiveness of our proposed model (Fig. 3D). The model had a higher AUC than clinical features, demonstrating its reliability for predicting the survival rates of HNSCC patients (Fig. 3E and 3F).
Survival forecasting and ROC curves for the additional datasets were illustrated in Fig. S2 and S3. 46; 47. The subgroup analysis revealed a statistically significant difference in survival rates between low-risk(higher survival rate) and high-risk patient groups, as illustrated in Fig 4A. Furthermore, we utilized PCA to examine the distribution patterns of the entire gene expression profiles, NRGs, NRlncRNAs, and 12 NRlncRNAs in the risk model. The distributions of the different risk groups were mixed and difficult to distinguish in Fig. 4B–4D. Yet, our research model suggests that there are notable divergences in the distributions across different risk categories (Fig. 4E). There are substantial variations between the distributions of low- and high-risk groups, affirming the utility of our model in identifying and quantifying risk distribution patterns. These findings suggest that the signature we built can accurately differentiate between different risk groups, emphasizing the model's clinical potential for patient risk stratification and tailored treatment plans.
Assessment of risk model
Based on the results of our study, we performed univariate Cox analysis to obtain the HR and 95% CI of the risk score, which were 1.384 and 1.284-1.492 (p < 0.001), as shown in Fig. 5A. Furthermore, multivariate Cox analysis revealed a HR of 1.325 and 95% CI of 1.222-1.436 (p < 0.001), indicating that the risk model developed from the 12 NRlncRNAs was independent of the clinicopathological features of HNSCC patients, as presented in Fig. 5B. The C-index and DCA demonstrated the impressive potential of the model as a reliable predictor of HNSCC prognosis, surpassing other clinical factors, as depicted in Fig. 5C and 5D. Furthermore, we formulated a nomogram to aid in estimating the survival probabilities of patients with HNSCC, as depicted in Fig 5E.
Estimation of the immune landscape
We first carried out GO analysis to explore the potential molecular mechanisms of the model, which revealed the participation of numerous immune-related biological processes (Fig. 6A; Table S1). We next performed KEGG enrichment analysis, and the correlations of 18 pathways were shown in Fig. 6B and Table S2. Utilizing established methodologies, we observed that the high-risk group exhibited increased abundance of mast cells, CD4+ T cells, and common lymphoid progenitor cells, while the low-risk group demonstrated higher levels of B cells, M0 macrophages, and CD8+ T cells (Fig 7A). Statistical differences were found in all parts of immune function except for the major histocompatibility complex class I, parainflammation, and response to type I interferon (Fig. 7B). Next, we compared the expression of 47 ICIs-related genes in both different groups, finding that gene expression in different groups had differences (Fig. 7C).
The model identified TP53, TTN, FAT1, CDKN2A, MUC16, PIK3CA, CSMD3, NOTCH1, SYNE1, and LRP1B as the top 10 mutated genes, as presented in Fig 8A and B. Patients in the high-risk group exhibited a significantly higher frequency of TP53 and TTN mutations in comparison to other mutations. Regarding the CSMD3 mutation levels, on the other hand, the exact reverse was seen. As anticipated, the high-risk group responded to immunotherapy more readily, showing that this necroptosis-based classifier score may be used for the prediction of TIDE and TMB (Fig. 8C and D). Survival analysis showed that different TMB and risk groups had statistically different survival rates (Fig. 8E and 8F). Therefore, we could screen suitable immunotherapeutic medications for HNSCC patients. We also discovered 27 commonly used antitumor drugs for different risk groups (Fig. S4).
Quantitative RT-PCR analysis
We aimed to validate the expression levels of five NRlncRNAs (JPX, AC079145.1, AC004943.2, MIR4435.2HG, and AL355574.1) in WSU-HN30 and SCC25 cells using qRT-PCR analysis. Our observations suggest that cancer cells upregulate the mRNA expression levels of the 5 NRlncRNAs, as depicted in Fig 9.