Multiomics and Machine Learning Identify Prognostic Immune Related Gene Signatures in Ovarian Cancer

Objective Ovarian cancer exhibits high heterogeneity, significant prognostic variability, and a lack of reliable prognostic biomarkers. This study aimed to construct and validate risk-associated gene signatures for ovarian cancer using multi-omics integration and machine learning techniques, thereby providing support for precise diagnosis and treatment. Methods Single-cell transcriptome data of ovarian cancer and bulk transcriptome data with clinical prognostic information were collected. Major cell types were identified via UMAP clustering. Weighted Gene Co-Expression Network Analysis (WGCNA) was used to screen modules strongly associated with risk traits. Prognostic genes were initially filtered through differential expression analysis and univariate Cox regression, followed by refinement using LASSO regression to obtain 14 risk-associated genes, which were then used to construct a prognostic risk signature model. Results Kaplan-Meier survival analysis showed a significant difference in survival rates between the high-risk and low-risk groups (p = 0.028). In subgroup analysis, the survival rate of the Cluster C1 group was significantly lower than that of the C2 group (p = 0.012). Receiver Operating Characteristic (ROC) curve analysis demonstrated that the area under the curve (AUC) values of the model for 1-year, 2-year, and 3-year survival prediction were 0.60, 0.66, and 0.64, respectively. After integrating the risk score with clinical factors into a nomogram, the calibration curve confirmed a high consistency between the predicted and actual survival outcomes. The immune score and stromal score of the high-risk group were significantly higher than those of the low-risk group, with high expression of immune marker genes such as SRGN and PTPRC. Drug sensitivity analysis revealed that the high-risk group was more sensitive to Foretinib and Pictilisib, while the low-risk group was more sensitive to Cediranib. Conclusion The 14-gene risk signature model constructed in this study exhibits stable prognostic predictive ability, can correlate with the state of the immune microenvironment, and guide personalized medication. It provides a reliable molecular tool and theoretical basis for precise prognostic evaluation and treatment option selection in ovarian cancer.