Plasma cell-free DNA 5-hydroxymethylcytosine signatures as predictive biomarkers for platinum resistance and prognosis in epithelial ovarian cancer

Platinum resistance is a major determinant of poor prognosis in epithelial ovarian cancer (EOC); however, no reliable pretreatment biomarkers are currently available to guide first-line therapy. This study aimed to develop a plasma cell-free DNA (cfDNA) 5-hydroxymethylcytosine (5hmC) signature for predicting platinum response and overall survival (OS). In this prospective pilot study, pretreatment plasma samples were collected from 63 patients with EOC scheduled to receive platinum-based chemotherapy. Based on follow-up outcomes, 20 patients were classified as drug-resistant (DR) and 43 as drug-sensitive (non-DR). For survival analysis, patients were stratified into short-term (n = 25) and long-term (n = 38) survivor groups using the median OS (36 months) as the cutoff. Patients were randomly assigned to training and validation sets cohorts in an 8:2 ratio. Differentially hydroxymethylated regions (DhMRs) were identified using 5hmC-Seal sequencing and subjected to recursive feature elimination for feature selection. Logistic regression models were subsequently constructed. Model performance was evaluated and compared with conventional biomarkers, including CA125 and CA19-9. Using a machine learning algorithm, eight 5hmC markers associated with platinum resistance were identified. In the validation set, the logistic model achieved a sensitivity of 75.0% and a specificity of 100.0% (AUC = 0.944). Eight 5hmC markers were identified for the prognosis prediction model, which achieved a sensitivity of 85.7% and a specificity of 80.0%, with an AUC of 0.914 in the validation set. The predictive performance of these models was superior to that of CA125 and CA19-9. Integration of the eight markers with CA125 and CA19-9 further improved predictive performance (AUC > 0.970). Downregulation of SLCO3A1 reversed cisplatin resistance, inhibited cell migration, and promoted cell proliferation, potentially via the Wnt/β-catenin signaling pathway. Our results demonstrate that plasma cfDNA 5hmC signatures provide a minimally invasive epigenetic read-out for the preliminarily prediction of drug resistance and prognosis in patients with EOC. SLCO3A1 may represent a potential therapeutic target associated with platinum resistance and poor prognosis in EOC.