Development and validation of an epigenetic score for homologous recombination deficiency based on genome-wide DNA methylation profiling in ovarian cancer

Abstract

Homologous recombination deficiency (HRD) is a key determinant of treatment sensitivity in high-grade serous ovarian carcinoma (HGSOC), yet current genomic scar-based assays capture the accumulated historical consequences of DNA damage rather than the dynamic functional state of the homologous recombination pathway. We developed an epigenetic classifier of HRD — the Epi-HRD Score — using LASSO regression on genome-wide DNA methylation data (Illumina HumanMethylation27; n = 548) from The Cancer Genome Atlas ovarian cancer cohort. The model identified a 189-CpG signature achieving an area under the ROC curve (AUC) of 0.854 on an independent test set, with stability confirmed by 20-repeat 5-fold cross-validation (mean AUC 0.861 ± 0.034). A compact 50-probe submodel retained 98.3% of full-model performance. Patients with high Epi-HRD Scores showed significantly longer overall survival (median 59.1 vs. 38.0 months; HR = 0.552; 95% CI 0.450–0.677; p = 7.19 × 10⁻⁹). In multivariate Cox analysis the score was co-linear with genomic HRD, supporting its interpretation as an epigenetic surrogate. Genome-wide methylation–expression analysis demonstrated significant coupling at all 20 top-ranked loci, and mediation analysis identified POLK — a Y-family translesion synthesis polymerase — as the strongest mediator (proportion mediated = 31.8%). A four-group concordance framework identified tumors with epigenetic but not genomic HRD (n = 34), potentially retaining epigenetic memory of a prior HRD state. External validation in an independent EPIC cohort showed preliminary cross-platform portability (AUC = 0.725). The Epi-HRD Score is a compact, cross-platform epigenetic measure that complements genomic scar-based HRD assessment.

Note:
Funding declaration: This study was supported by the Natural Science Foundation of Shanghai (Grant No. 25ZR1401319).

Conflict of Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
<

Keywords: Homologous recombination deficiency, DNA methylation, ovarian cancer, epigenetic biomarker, LASSO regression, PARP inhibitor, prognostic signature