Identification and Validation of a Novel Anoikis-Related Gene Signature for Predicting Survival in Patients With Serous Ovarian Cancer

Abstract

Background Ovarian cancer is an extremely deadly gynecological malignancy, with a 5-year survival rate below 30%. Among the different histological subtypes, serous ovarian cancer (SOC) is the most common. Anoikis significantly contributes to the progression of ovarian cancer. Therefore, identifying an anoikis-related signature that can serve as potential prognostic predictors for SOC is of great significance. Methods We intersected 308 anoikis-related genes (ARGs) and identified those significantly associated with SOC prognosis using univariate Cox regression. A LASSO Cox regression model was constructed and evaluated using Kaplan-Meier and receiver operating characteristic (ROC) analyses in TCGA (The Cancer Genome Atlas) and GSE26193 cohorts. We conducted quantitative real-time polymerase chain reaction (qPCR) to assess mRNA levels and applied bioinformatics to investigate the correlation between risk groups and gene expression, mutations, pathways, tumor immune microenvironment (TIME), and drug sensitivity in SOC. Results Among 308 ARGs, 28 were significantly associated with SOC prognosis. A 13-gene prognostic model was established through LASSO Cox regression in TCGA cohort. High-risk group had poorer prognosis than low-risk group (median overall survival (mOS): 34.2 vs. 57.1 months, hazard ratio (HR): 2.590, 95% confidence interval (CI): 0.159 - 6.00, P < 0.001). The area under the curve (AUC) values of 0.63, 0.65, and 0.74 reflected the predictive performance for 3-, 5-, and 8-year overall survival (OS) in GSE26193 validation cohort. Functional enrichment, pathway analysis, and TIME analysis identified distinct characteristics between risk groups. Drug sensitivity analysis revealed potential drug advantages for each group. Furthermore, qPCR validation once again confirmed the effectiveness of the risk model in SOC patients. Conclusions We developed and validated a robust ARG model, which could be used to predict OS in SOC patients. By systematically analyzing the correlation between the risk score of the ARGs signature model and various patterns, including the TIME and drug sensitivity, our findings suggest that this prognostic model contributes to the advancement of personalized and precise therapeutic strategies. Nevertheless, further validation studies and investigations into the underlying mechanisms are warranted.