The Development of Three-DNA Methylation Signature as a Novel Prognostic Biomarker in Patients with Colorectal Cancer

Abstract

Aims. The prognosis of colorectal cancer (CRC) remains poor. This study aimed to develop and validate DNA methylation-based signature model to predict overall survival of CRC patients. Methods. The methylation array data of CRC patients were retrieved from The Cancer Genome Atlas (TCGA) database. These patients were divided into training and validation datasets. A risk score model was established based on Kaplan-Meier and multivariate Cox regression analysis of training cohort and tested in validation cohort. Results. Among total 14,626 DNA methylation candidate markers, we found that a three-DNA methylation signature (NR1H2, SCRIB, and UACA) was significantly associated with overall survival of CRC patients. Subgroup analysis indicated that this signature could predict overall survival of CRC patients regardless of age and gender. Conclusions. We established a prognostic model consisted of 3-DNA methylation sites, which could be used as potential biomarker to evaluate the prognosis of CRC patients.

DNA methylation-based biomarkers and prediction models for the survival of patients with colorectal cancer: systematic review and external validation study

“…7-12 14 28 30 The equation to calculate prognostic score, the distribution of each prognostic score in our validation cohort, and the performance of these models reported by original development studies are shown in Supplementary Table 6. Five prognostic models from four studies 8-10 14 were independently associated with OS (Supplementary Table 7). The relationships between the prognostic scores and survival were largely linear (Figure 2A).…”

“…Only one study handled missing values properly by multiple imputations, 9 while the remaining studies either excluded patients with missing values or did not report this information. Four out of nine studies (44%) did not take measures to account for model overfitting, 7 8 10 14 and three of the nine studies selected CpGs simply based on the p -values in univariable and multivariable analyses without considering established clinical knowledge. 8 14 28…”

“…Four out of nine studies (44%) did not take measures to account for model overfitting, 7 8 10 14 and three of the nine studies selected CpGs simply based on the p -values in univariable and multivariable analyses without considering established clinical knowledge. 8 14 28…”

“…6 Accumulating evidence supports the prognostic relevance of various DNA methylation biomarkers in patients with CRC, 3 and several prognostic models based on prognostically-relevant CpGs have been developed. [7][8][9][10][11][12][13][14][15] However, so far, none of these potential prognostic markers or models have been translated to clinical practice, hampered by a lack of validation analyses in large independent patient cohorts, 3 which is a huge problem in cancer biomarker research in general. Innumerable biomarkers are published each year, but only very few ever make it to clinical practice.…”

A Multi-Omics Overview of Colorectal Cancer to Address Mechanisms of Disease, Metastasis, Patient Disparities and Outcomes

“…This classifier included probes in BRD4, FBXL18, KCNQ1 and MET, among others, and the DNA methylation score of this panel was positively correlated with patient survival [61]. (3) Gong et al mined TCGA CRC DNA methylation data to compile a three-gene (NR1H2, SCRIB and UACA) DNA methylation signature that correlated with patient OS irrespective of patient age and gender [62]. (4) Xie et al [63] used tumor-derived cfDNA from blood plasma to evaluate the performance of a 14-gene panel (ARHGEF4, AV3, BMP3, CHST2, DOCK10, IKZF1, LRRC4, NDRG4, OPLAH, PPP2R5C, PDGFD, QKI, SFMBT2 and ZNF625).…”