Abstract
INTRODUCTION: Epithelial ovarian cancer (EOC) represents a significant health challenge, with high-grade serous ovarian cancer (HGSOC) being the most common subtype. Early detection is hindered by nonspecific symptoms, leading to late-stage diagnoses and poor survival rates. Biomarkers are crucial for early diagnosis and personalized treatment OBJECTIVE: Our goal was to develop a robust statistical procedure to identify a set of differentially methylated probes (DMPs) that would allow differentiation between HGSOC and benign ovarian tumors. METHODOLOGY: Using the Infinium EPIC Methylation array, we analyzed the methylation profiles of 48 ovarian samples diagnosed with HGSOC, borderline ovarian tumors, or benign ovarian disease. Through a multi-step statistical procedure combining univariate and multivariate logistic regression models, we aimed to identify CpG sites of interest. RESULTS AND CONCLUSIONS: We discovered 21 DMPs and developed a predictive model validated in two independent cohorts. Our model, using a distance-to-centroid approach, accurately distinguished between benign and malignant disease. This model can potentially be used in other types of sample material. Moreover, the strategy of the model development and validation can also be used in other disease contexts for diagnostic purposes.