AUTHOR=Xu Mingyue , Yuan Lijun , Wang Yan , Chen Shuo , Zhang Lin , Zhang Xipeng 

TITLE=Integrative Analysis of DNA Methylation and Gene Expression Profiles Identifies Colorectal Cancer-Related Diagnostic Biomarkers

JOURNAL=Pathology and Oncology Research

VOLUME=Volume 27 - 2021

YEAR=2021

URL=https://www.por-journal.com/journals/pathology-and-oncology-research/articles/10.3389/pore.2021.1609784

DOI=10.3389/pore.2021.1609784

ISSN=1532-2807

ABSTRACT=Background: Colorectal cancer (CRC) is a common human malignancy worldwide. The prognosis of patients is largely frustrated by delayed or misdiagnosis. DNA methylation alterations have been previously proved to be involved in CRC carcinogenesis. 
Methods: In this study, we proposed to identify CRC related diagnostic biomarkers by integrated analyzing DNA methylation and gene expression profiles. TCGA-COAD datasets downloaded from the Cancer Genome Atlas (TCGA) were used as training set to screen differential expression genes (DEGs) and methylation CpG sites (dmCpGs) in CRC samples. A logistic regression model was constructed based the hyper-methylated CpG sites which was located in down-regulated genes for CRC diagnosis. Another two independent datasets from Gene Expression Omnibus (GEO) were used as testing set to evaluate performance of the model in CRC diagnosis. 
Results: We found that CpG island methylator phenotype (CIMP) was a potential signature of poor prognosis by dividing CRC samples into CIMP and noCIMP group based on a set of CpG sites with methylation standard deviation (sd) > 0.2 among CRC samples and low methylation levels (mean β < 0.05) in adjacent samples. Hyper-methylated CpGs tended to be more closed to CpG island (CGI) and transcription start site (TSS) relative to hypo-methylated CpGs (p-value < 0.05, Fisher exact test). A logistic regression model was finally constructed based on two hyper-methylated CpGs, which area under receiver operating characteristic curve achieved 0.98 in the training set, and 0.85 and 0.95 in the two independent testing sets. 
Conclusions: In conclusion, our study identified promising DNA methylation biomarkers for CRC diagnosis.