Artificial intelligence for single-omics in ovarian cancer: a methodological review

Ovarian cancer is a leading cause of cancer-related mortality among women, with poor prognosis and limited survival in advanced stages. The integration of artificial intelligence with omics data offers new opportunities to enhance the diagnosis, prognosis, and treatment of this disease. This narrative review synthesizes evidence from 14 studies published between 2021 and 2024 that applied artificial intelligence to genomic, transcriptomic, metabolomic, micro-biomic, and epigenomic data sets in patients with epithelial ovarian cancer. These studies explored artificial intelligence models for disease detection, chemotherapy response prediction, and genetic risk stratification. Despite promising results (eg, high classification accuracy and area under the curve values in some models), significant limitations were observed, including small sample sizes, retrospective and single-center designs, and inconsistent use of validation data sets. The review highlights critical methodological considerations such as data preprocessing, normalization, and feature selection, which substantially influence model performance and reproducibility. Although classification models (eg, deep learning, random forest, and support vector machines) were most commonly used, regression approaches were less frequent and under-used, despite their value for modeling continuous outcomes such as survival time. Overall, artificial intelligence-based approaches demonstrate great potential for advancing personalized medicine in ovarian cancer. However, future research must prioritize larger, multi-center, prospective studies with robust validation strategies and improved model interpretability to enable clinical implementation.