DEVELOPMENT OF A CLINICAL AND GENOMIC DATA-BASED MACHINE LEARNING MODEL FOR PREDICTING THE PROGNOSIS OF HIGH-GRADE SEROUS OVARIAN CANCER

Abstract

High-Grade Serous Ovarian Cancer (HGSOC) is among the most aggressive gynecologic malignancies with increasing mortality trends in several developing countries. Conventional prognosis assessment based solely on clinical staging often fails to reflect the molecular complexity of the disease. This study proposes machine learning models integrating clinical and genomic features to predict HGSOC prognosis. Data from 617 patients obtained from The Cancer Genome Atlas were preprocessed using imputation, feature selection, class balancing with Synthetic Minority Oversampling Technique, and hyperparameter optimization. Random Forest and Extreme Gradient Boosting were evaluated using accuracy, precision, recall, and F1-score. Both optimized models achieved 0.86 accuracy and 0.85 weighted F1-score. Feature contribution analysis using SHapley Additive exPlanations indicated that clinical stage, histology code, and diagnosis year were dominant predictors. The findings demonstrate that integrating structured clinical and genomic data yields reliable prognostic performance while maintaining interpretability.