Copy number signatures in cervical samples enable early detection of high-grade serous ovarian carcinoma

BackgroundThe lack of effective screening methods for the detection of ovarian cancer leads to late diagnosis and poor outcomes. High-grade serous ovarian carcinoma (HGSC) is strongly dominated by copy number alterations (CNAs), resulting from underlying mutational processes. Several studies have reported the detection of ovarian carcinoma-derived somatic mutations in cervical samples, uterine lavages, and plasma, which has brought hope for the development of new biomarkers for early detection.MethodsUsing 212 cervical samples from 128 women (58 HGSC, 22 healthy BRCA1/2 mutation carriers and 48 women with benign gynecological conditions), we interrogated CNAs from low-coverage whole-genome sequencing (WGS) data from cervical samples collected at surgery and at multiple time points up to 94 months before surgery. A machine-learning-based classifier was used to identify patterns of CNAs and to construct the HCsig predictor.ResultsUsing HCsig, 79% of HGSC patients had at least one positive cervical sample before surgery (up to 65 months); 91% for women with stage I-II disease (up to 27 months), and 78% for women with stage III-IV disease (up to 65 months). Notably, 90% of the BRCA1/2-mutated tumors were detectable in at least one pre-diagnostic cervical sample (up to 54 months), compared to 76% of the BRCA1/2 wildtype tumors (up to 51 months). Using an independent dataset of 172 samples for validation, HGSC was detected with a specificity of 94% and sensitivity of 76% (AUC=0.83); importantly, including high sensitivity for early-stage cancers.ConclusionsOur findings show that HGSC-derived copy number features of HCsig can detect HGSC with high performance in archival cervical samples collected several years before the development of symptoms and eventual diagnosis. Secondary prevention based on earlier-stage detection using HCsig addresses the unmet need for HGSC screening and detection.