Construction and validation of a metabolic-related genes prognostic model for oral squamous cell carcinoma based on bioinformatics

BMC Medical Genomics

Research

Jinghua Liu¹  Chenxi Ma²  Jingfei Zhang³  Han Qin³  Xiuyan Hao⁴  Zhi Wang⁴  Chao Zhu⁴  Xiujuan Liu⁴  Zhen Cai⁴  Ling Li⁴ 

[1] Department of Hepatobiliary Surgery and Minimally Invasive Institute of Digestive Surgery and Prof. Cai’s Laboratory, Linyi People’s Hospital, Shandong University, 264000, Linyi, Shandong, China;Department of Human Microbiome, School and Hospital of Stomatology, Shandong Provincial Key Laboratory of Oral Tissue Regeneration, Shandong Engineering Laboratory for Dental Materials and Oral Tissue Regeneration, Shandong University, 250000, Jinan, Shandong, China;Department of Stomatology, Binzhou Medical University, 264000, Yantai, Shandong, China;Department of Stomatology, Linyi People’s Hospital, 276000, Linyi, Shandong, China;

关键词: Metabolic studies;    Oral squamous cell carcinoma;    Prognosis;    Nomogram;    Bioinformatics;   

DOI  :  10.1186/s12920-022-01417-3

 received in 2022-07-12, accepted in 2022-12-13,  发布年份 2022

来源: Springer

【 摘 要 】

BackgroundOral squamous cell carcinoma (OSCC) accounts for a frequently-occurring head and neck cancer, which is characterized by high rates of morbidity and mortality. Metabolism-related genes (MRGs) show close association with OSCC development, metastasis and progression, so we constructed an MRGs-based OSCC prognosis model for evaluating OSCC prognostic outcome.MethodsThis work obtained gene expression profile as well as the relevant clinical information from the The Cancer Genome Atlas (TCGA) database, determined the MRGs related to OSCC by difference analysis, screened the prognosis-related MRGs by performing univariate Cox analysis, and used such identified MRGs for constructing the OSCC prognosis prediction model through Lasso-Cox regression. Besides, we validated the model with the GSE41613 dataset based on Gene Expression Omnibus (GEO) database.ResultsThe present work screened 317 differentially expressed MRGs from the database, identified 12 OSCC prognostic MRGs through univariate Cox regression, and then established a clinical prognostic model composed of 11 MRGs by Lasso-Cox analysis. Based on the optimal risk score threshold, cases were classified as low- or high-risk group. As suggested by Kaplan–Meier (KM) analysis, survival rate was obviously different between the two groups in the TCGA training set (P < 0.001). According to subsequent univariate and multivariate Cox regression, risk score served as the factor to predict prognosis relative to additional clinical features (P < 0.001). Besides, area under ROC curve (AUC) values for patient survival at 1, 3 and 5 years were determined as 0.63, 0.70, and 0.76, separately, indicating that the prognostic model has good predictive accuracy. Then, we validated this clinical prognostic model using GSE41613. To enhance our model prediction accuracy, age, gender, risk score together with TNM stage were incorporated in a nomogram. As indicated by results of ROC curve and calibration curve analyses, the as-constructed nomogram had enhanced prediction accuracy compared with clinicopathological features alone, besides, combining clinicopathological characteristics with risk score contributed to predicting patient prognosis and guiding clinical decision-making.ConclusionIn this study, 11 MRGs prognostic models based on TCGA database showed superior predictive performance and had a certain clinical application prospect in guiding individualized.

【 授权许可】

CC BY   
© The Author(s) 2022

【 图 表 】

Fig. 3

Fig. 1

Fig. 6

Fig. 5

Fig. 1

Fig.5

Fig. 1