Early Diagnosis and Prognostic Prediction of Colorectal Cancer through Plasma Methylation Regions

Abstract The methylation of plasma cell-free DNA (cfDNA) has emerged as a valuable diagnostic and prognostic biomarker in various cancers including colorectal cancer (CRC). Currently, there are no biomarkers that serve simultaneously for early diagnosis and prognostic prediction in CRC patients. Herein, we developed a plasma panel (27 DMRs, differential methylated regions) and validated its superior performance across CRC diagnosis and prognosis prediction in an independent cohort. We first conducted a preliminary screening of 119 CRC tissue samples to identify CRC-specific methylation features. Subsequently, a CRC-specific methylation panel was developed by further filtering 161 plasma samples. Then machine learning algorithms were applied to develop diagnosis and prognosis models using cfDNA samples from 51 CRC patients and 33 normal controls. The diagnosis model was tested in a cohort consisting of 30 CRC, 37 advanced adenoma (AA), and 14 healthy plasma samples, independently validated in a cohort consisting of 18 CRC, 91 NAA, 23 AA and 34 healthy plasma samples. In the tissue external validation cohort (GSE48684), the cfDNA methylation diagnosis model conducted with the panel, have the area under the curve (AUC) reached 0.983, and for the plasma cfDNA model in the external validation cohort, the sensitivities for NAA, AA and CRC 0 -Ⅱ are 48.4%. 52.2% and 66.7% respectively, with a specificity of 88%. Additionally, the panel was applied to patient staging and metastasis, performing well in predicting CRC distant metastasis (AUC = 0.955) and prognosis (AUC = 0.867). Using normal samples as control, the changes in methylation score in both tissue and plasma were consistent across different lesions, although the degree of alterations varied with severity. The methylation scores vary between paired tissue and blood samples, suggesting distinct mechanisms of migration from tumor tissue to blood for the 27 DMRs. Together, Our cfDNA methylation models based on 27 DMRs can identify different stages of CRC and predict metastasis and prognosis, ultimately enabling early intervention and risk stratification for CRC patients.