Validation of novel DNA methylation markers in cervical precancer and cancer

Abstract

We have recently identified, using a genome-wide approach, new methylation markers which were evaluated among various cervical intraepithelial neoplasia (CIN) grades and cervical cancer. We herein validate the methylated state of these genes in independent study populations, based on histology ascertained outcomes regardless of human papillomavirus status. CA10, DPP10, FMN2 and HAS1 (discovery set: 54 normal, 50 CIN1, 40 CIN2, 42 CIN3) were evaluated by targeted bisulfite next generation sequencing (NGS) (Illumina MiSeq platform) in 258 (training set: 100 normal, 50 CIN1, 50 CIN2, 50 CIN3, 8 cancers) and 373 (validation set: 100 normal, 57 CIN1, 61 CIN2, 53 CIN3, 102 cancers) physician-collected samples (PreservCyt). Using targeted amplification NGS data from the training set for 94 normal and eight cancer samples, we calculated for each gene the median methylation value. These were summed and normalized to compute a four-gene Marker Polygenic Score (MPS). We compared the relationship between MPS and progression from normal through CIN grades and cancer, separately in the training and validation sets, and tested its clinical performance via receiver-operating characteristic curves. MPS increased with increasing CIN grade, and accurately predicted cervical cancer in the training (area under the curve, AUC = 0.9950) and validation (AUC = 0.9337) sets, comparing normal to cancer. Using the highest threshold of 100% specificity, sensitivity for detection of cervical cancer was 67.7%; whereas reducing specificity to 95% increased sensitivity to 84.3%. Further evaluation of these biomarkers is warranted in prospective studies.