Methodological Innovation and Secondary Analysis: Integrating DNA Methylation Signatures into Structural Equation Modeling for Ovarian Cancer Risk and Complication Prediction

Abstract Background: Often referred to as the ‘silent killer’, ovarian cancer remains a leading cause of gynaecological cancer deaths. Although genetic markers like BRCA 1/2 provide some risk assessment, they are not sufficient alone for the prevention, detection or treatment of ovarian cancer. Recent evidence highlights the importance of epigenetic regulations, specifically DNA methylation, in modifying the disease development treatment response. Primary Objective: To develop and validate a novel structural equation modelling (SEM) framework that integrates DNA methylation signatures for investigating epigenetic mechanisms in ovarian cancer. Secondary Objectives: (1) To identify co-methylated regions (CMRs) that act as latent variables mediating the relationship between genetic risk and clinical outcomes. (2) To evaluate the predictive performance of methylation-informed SEM models compared to traditional risk predictors. (3) To explore the biological pathways enriched in differentially methylated regions (DMRs) associated with BRCA status and treatment response. Methods: Whole-genome bisulphite sequencing was used in this study to integrate DNA methylation data into SEM for the investigation of epigenetic mechanisms in ovarian cancer. Latent methylation variables were derived from CMRs, and penalised regression was utilised to identify methylation mediators that link genetic risk to clinical outcomes, providing insights into cancer pathogenesis. Results: One thousand two hundred and forty seven DMRs associated with BRCA were identified, including enhancer-linked regions enriched for DNA repair and chromatin pathways. SEM revealed DNA methylation mediated 58% of progression-free survival variance and 68% of genetic risk on platinum resistance. Key CMRs near WNT3 and RARB explained major BRCA -related effects. Models integrating methylation outperformed traditional predictors (area under the curve = 0.82). Conclusion: These findings demonstrate the value of DNA methylation signatures as biomarkers for clinical diagnosis and disease monitoring. Future research will seek to broaden their applicability across different populations and incorporate them into standard clinical operations.