Machine Learning-Based Signature for Predicting Prognosis and Drug Sensitivity in Ovarian Cancer With Macrophage M2-Related Genes

Background: Ovarian cancer is the third most prevalent gynecological malignancy globally. M2 macrophages play crucial roles in promoting angiogenesis, cancer cell proliferation, metastasis, and immunosuppression.

Methods: We identified markers associated with M2 macrophages using weighted gene co-expression network analysis. A machine learning approach, encompassing ten algorithms, was employed to construct a macrophage M2-related signature (MRS) based on data from TCGA, GSE14764, and GSE140082 datasets. The predictive value of MRS for immunotherapy response was assessed using immunophenoscore, TIDE score, tumor mutational burden (TMB) score, and immune escape score.

Results: The optimal MRS, developed using the lasso algorithm, emerged as an independent risk factor and demonstrated robust performance in predicting overall survival in ovarian cancer patients. The C-index of our MRS surpassed that of clinical stage, tumor grade, and several established prognostic signatures. Patients with lower risk score exhibited higher ESTIMATE score, increased levels of immune cells, elevated PDI and CTLA4 immunophenoscore, higher TMB score, lower TIDE score, reduced immune escape score, and decreased IC50 values for certain drugs. The nomogram for survival prediction showed significant potential for clinical application in forecasting 1-, 3-, and 5-year overall survival rates in ovarian cancer patients.

Conclusion: Our study developed a stable MRS for ovarian cancer, which serves as a valuable indicator for predicting prognosis and drug sensitivity in this disease. Further prospective studies should be performed to further explore the role of MRS in predicting the clinical outcome and immunotherapy benefits of ovarian cancer patients.

Trial Registration: ClinicalTrials.gov identifier: NCT02108652