Identification of a novel gene expression signature as prognostic biomarker for human papillomavirus negative head and neck carcinoma

Abstract

AbstractPurposeHead and neck squamous cell carcinomas (HNSCCs) are heterogeneous and complex malignancies. The present study aimed to explore potential molecular classifications and establish promising prognostic candidate biomarkers for human papillomavirus (HPV)‐negative HNSCC.MethodsWe performed unsupervised clustering in HPV‐negative patients from the TCGA‐HNSCC database and identified molecular subclusters based on HNSCC prognosis. Weighted gene co‐expression network analysis was used to identify cluster‐related gene modules and identify reliable candidate genes. Biomarker signature was established from these candidate genes via 10‐fold cross validation in a least absolute shrinkage and selection operator (LASSO) Cox regression model and validated in The Cancer Genome Atlas Program (TCGA) cohort and four external Gene Expression Omnibus (GEO) validation cohorts.ResultsWe identified three molecular subclusters (C1–C3), and tumour microenvironment and survival evaluation results showed that C3 had high immune cell infiltration and better prognosis. However, C1 and C2 had low immune cell infiltration, and C1 had a worse survival outcome. A 20‐gene signature was established from the input candidate genes via 10‐fold cross validation in the LASSO Cox regression model. The signature was first validated in a training cohort, followed by a testing cohort and the entire cohort. A total of 347 patients from four external GEO validation cohorts indicated great performance of the signature in predicting outcomes.ConclusionCompared with HPV‐positive patients, those with HPV‐negative HNSCC had an immunosuppressed tumour microenvironment. The 20‐gene signature was developed based on the characteristic unsupervised clustering molecular subclusters to predict survival outcomes in patients with HPV‐negative HNSCC.