Development of CSOARG: a single-cell and multi-omics-based machine learning model for ovarian cancer prognosis and drug response prediction.

OBJECTIVE: Ovarian cancer is the most deadly gynaecological malignancy. This study aims to generate a predictive model for prognosis and therapeutic responses in ovarian cancer using defined specific genes. METHODS: The cellular senescence-associated gene sets and the ovarian aging-associated gene sets from the TCGA and GEO databases were analyzed using Cox regression with LASSO approach and employed to construct a prognostic model of Cellular Senescence and Ovarian Aging-Related Genes (CSOARG). Immunology analysis, functional enrichment, single-cell analysis, and therapeutic responses of ovarian cancer were conducted using the data from public databases. A machine learning model based on the expression levels of prognostic genes combined with clinical features was developed to predict the five-year overall survival. Patients with high- and low-risk scores were separated by the median risk score. Defined genes were verified by qRT-PCR and Western blot. The cellular behavior was evaluated by CCK-8, migration, and wound-healing assays. RESULTS: After a series of calculations, an 8-gene CSOARG model was generated. CSOARG was correlated with genomic instability that harbored homologous recombination deficiency. The area under the curve (AUC) for 5-year overall survival was 0.68. Patients in the high-risk score group had a higher IC(50) of chemotherapeutic and targeted therapeutical agents, worse responses to chemotherapy and immunotherapy, and exhibited a poor prognosis. A hub gene WNK1 was validated and acted as an oncogene affecting ovarian cancer cell viability and migration. CONCLUSIONS: These findings demonstrate that a novel CSOARG model can effectively predict the prognosis and therapeutical responses of patients with ovarian cancer, which may assist clinicians in implementing better practices.