Machine learning analysis of oxidative stress-related phenotypes for specific gene screening in ovarian cancer.

Abstract

Oxidative stress serves a crucial role in tumor development. However, the relationship between ovarian cancer and oxidative stress remains unknown. We aimed to create an oxidative stress-related prognostic signature to enhance the prognosis prediction of CC patients using bioinformatics. The genes differentially expressed and associated with oxidative stress were extracted with the help of "limma" packages. The model for prognosis was created using Multivariate Cox regression analysis to determine the risk related to the genes related to oxidative stress. Patients were categorized as low-risk or high-risk based on the median score. The receiver operation characteristic (ROC) and survival curves were used to evaluate the predictive effect of the prognostic signature. We utilized quantitative real-time PCR to assess the expression levels of key genes associated with oxidative stress in ovarian cancer cell lines (SKOV3, OVCAR3, and HeyA8) and normal ovarian epithelial cells (HOSEpiC). A signature comprising seven genes associated with oxidative stress was developed to prognosticate patients with ovarian cancer. Overall survival (OS) of the patient having CC was determined using Kaplan-Meier analysis. It was found that patient with a higher risk score had lower OS than the low-risk score. The signature of genes associated with oxidative stress was found to be independently prognostic for 1, 2, and 3 years. Further research found that the expression levels of nine hub genes had a strong association with patient outcomes. Our analysis revealed a higher expression of CX3CR1 in ovarian cancer cell lines compared with normal cells. To deploy a novel oxidative stress-related prognostic signature as an independent biomarker in cervical cancer, we developed and validated it.