A novel 4-gene signature model simultaneously predicting malignant risk of oral potentially malignant disorders and oral squamous cell carcinoma prognosis. (September 2021)

• Record Type: Journal Article
• Authors: Zhang, Xinyue; Yang, Miao; Liu, Yangfan; Liu, Hailong; Yang, Jin; Luo, Jingjing; Zhou, Hongmei
• Is Part Of: Archives of oral biology. Volume 129(2021)
• Publication Date: 2021-09
• Subjects: OSCC oral squamous cell carcinoma -- OPMDs oral potentially malignant disorders -- GEO Gene Expression Omnibus -- DEGs differentially expressed genes -- SVM support vector machine -- OCFS oral cancer-free survival -- GEP gene expression profiling -- RMA robust multi-array average -- TPR true positive rate -- FPR false positive rate -- AUC area under the curve -- IOD integral optical density; Oral squamous cell carcinoma -- Oral potentially malignant disorders -- Support vector machine -- Early diagnosis -- Prognosis -- Survival analysis; Mouth -- Periodicals; Mouth -- Diseases -- Periodicals; Dentistry -- Periodicals; Electronic journals; 617.6005
• DOI: 10.1016/j.archoralbio.2021.105203
• Languages: English
• ISSNs: 0003-9969

Abstract

Highlights: A novel 4-gene signature model with high accuracy was built by SVM. The model could simultaneously predict malignant risk of OPMDs and OSCC prognosis. The reliability of 4 gene signatures have been confirmed in human samples by IHC. The model provided an alternative for early diagnosis and prognosis of OSCC. Abstract: Objective: Oral squamous cell carcinoma (OSCC) is often diagnosed at late stage with a poor prognosis. The study hereunder aimed to construct a multi-gene model to simultaneously promote early diagnosis of OSCC by evaluating malignant risk of oral potentially malignant disorders (OPMDs) and predict prognosis. Materials and methods: 3 GEO datasets including OPMDs and OSCC samples were obtained for overlapping differentially expressed genes (DEGs) being screened. The predictive model was built with optimal DEGs by SVM algorithm, estimated by receiver operator characteristic curves and validated for double prediction via oral cancer-free survival (for malignant risk of OPMDs) and overall survival time (for OSCC) analysis respectively compared to other models. The protein expression of biomarkers in the model was validated in human samples by immunohistochemistry. Results: A novel predictive model of 4-gene signature was built based on 12 common DEGs revealed from 3 GEO datasets. It could well distinguish OSCC from OPMDs and normal tissues. Both oral cancer-free survival and overall survival time analysis were significantly poorer in high-risk patients than in low-risk ones in Kaplan Meier survival curve respectively. The protein expression of biomarkers in OSCC was with significant difference compared to normal and OPMDs. Conclusions: The novel 4-gene signature model presents strong ability in simultaneous prediction of the malignant risk of OPMDs and OSCC progression, potentially benefiting both the early diagnosis and therapeutic outcomes of OSCC.