Identification of a Six-Gene Signature for Predicting the Overall Survival of Cervical Cancer Patients.

Background: Although the incidence of cervical cancer has decreased in recent decades with the development of human papillomavirus vaccines and cancer screening, cervical cancer remains one of the leading causes of cancer-related death worldwide. Identifying potential biomarkers for cervical cancer treatment and prognosis prediction is necessary.

Methods: Samples with mRNA sequencing, copy number variant, single nucleotide polymorphism and clinical follow-up data were downloaded from The Cancer Genome Atlas database and randomly divided into a training dataset (N=146) and a test dataset (N=147). We selected and identified a prognostic gene set and mutated gene set and then integrated the two gene sets with the random survival forest algorithm and constructed a prognostic signature. External validation and immunohistochemical staining were also performed.

Results: We obtained 1416 differentially expressed prognosis-related genes, 624 genes with copy number amplification, 1038 genes with copy number deletion, and 163 significantly mutated genes. A total of 75 candidate genes were obtained after overlapping the differentially expressed genes and the genes with genomic variations. Subsequently, we obtained six characteristic genes through the random survival forest algorithm. The results showed that high expression of , and and low expression of and were associated with a poor prognosis in cervical cancer patients. We constructed a six-gene signature that can separate cervical cancer patients according to their different overall survival rates, and it showed robust performance for predicting survival (training set: ˂ 0.001, AUC = 0.82; testing set: ˂ 0.01, AUC = 0.59).

Conclusion: Our study identified a novel six-gene signature and nomogram for predicting the overall survival of cervical cancer patients, which may be beneficial for clinical decision-making for individualized treatment.