Abstract
Advanced adenoma (AA) holds a significantly increased risk for progression to colorectal cancer (CRC), and we developed a noninvasive DNA methylation prediction model to monitor the risk of AA progression to CRC. We analyzed the differential methylation markers between 53 normal mucosa and 138 CRC tissues, as well as those in cfDNA (cell-free DNA) between 59 AA and 68 early-stage CRC patients. We screened the overlapping markers between tissue DNA and cfDNA for model variables and optimized the selected variables. Then, we established a cfDNA methylation prediction model (SDMBP model) containing seven methylation markers that can effectively discriminate early-stage CRC and AA in the training and validation cohorts, and the AUC (area under the curve) reached 0.979 and 0.918, respectively. Our model also reached high precision (AUC=0.938) in detecting advanced CRC (stage III/IV) and presented better performance than serum CEA and CA199 in screening CRC. The cd-score of the SDMBP model could also robustly predict the TNM stage of CRC. Overall, our SDMBP model can monitor the malignant progression from AA to CRC, and may provide a noninvasive monitoring method for high-risk populations with AA.