Abstract Introduction: This study is focused on early detection of ovarian cancer using serum-derived microRNA (miRNA) expression profiles as non-invasive biomarker source. Ovarian cancer is among the most fatal gynecologic diseases, mainly because of its silent progression and late-stage diagnoses. Traditional screening methods, like CA-125, can be useful in detecting other cancers such as breast cancer. However, its low specificity, which results in significant false-positive rates due to rise in benign diseases including endometriosis, menstruation, and even a normal pregnancy, means that it cannot be considered as a reliable biomarker for ovarian cancer. As a result, we can see the urgency of the need to develop sensitive and specific approaches that can accurately identify cancer in its early stages. Method: We employed Machine Learning and Deep Learning techniques to analyze high-dimensional miRNA datasets from publicly available repositories of National Center for Biotechnology Information (NCBI). A high-dimensional dataset is a dataset with many features or variables compared to the number of observations. A situation where the feature count exceeds 10% of the sample size can challenge a machine learning model's performance. The datasets we used, for instance, contain 700 miRNA blood serum samples of ovarian cancer with 2550 features each. The key innovation in this work lies in its use of blood-based miRNA signatures, which can be collected through a simple, non-invasive blood draw. Compared to more invasive diagnostic procedures like images of tissue biopsy, this approach offers a scalable and patient-friendly alternative with support of earlier works in this area without compromising accuracy. This can be a huge step forward to opening the door of integrating miRNA screening into routine annual health checkups for women at elevated risk of ovarian or related cancers. Results: Our analysis demonstrates that cancerous cases can be accurately identified with 99% accuracy and an F1-score of 0. 99. This high level of precision validates serum miRNA signatures as powerful and reliable discriminative markers. Our method also gave promising results on distinguishing early-stage from late-stage ovarian cancer cases, a critical achievement given the dramatic difference in survival outcomes between these groups. A notable breakthrough is the achievement of high classification accuracy even in Stage I cases. These findings demonstrate the potential of our framework to address a critical gap in ovarian cancer diagnostics. Conclusion: Our study provides promising evidence that pairing blood serum miRNA profiling with advanced computational analysis offers a robust and specific method for the early detection and staging of ovarian cancer. Consequently, these results establish a strong foundation for advancing this non-invasive technique into larger clinical validation studies to confirm its potential clinical utility. Citation Format: Zahra Saghaie Dehkordi, Christine Richardson, Taghi Taghi Mostafavi. Ovarian cancer stage detection study using blood serum miRNA abstract. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Ovarian Cancer Research; 2025 Sep 19-21; Denver, CO. Philadelphia (PA): AACR; Cancer Res 2025;85 (18Suppl): Abstract nr B014.
Dehkordi et al. (Fri,) studied this question.