Colorectal cancer is the second most deadly malignancy in the United States. However, the currently screening options had their limitation. Novel biomarkers for colorectal cancer detections are necessary to reduce the mortality. The clinical information, mRNA expression levels and DNA methylation information of colorectal cancer were downloaded from TCGA. The patients were separated into training group and testing group based on their platforms for DNA methylation. Beta values of DNA methylation from tumor tissues and normal tissues were utilized to figure out the position that were differentially methylated. The expression levels of mRNA of thirteen genes, whose CpG islands were differentially methylated, were extracted from the RNA-Seq results from TCGA. The probabilities whether the mRNA was differentially expressed between tumor and normal samples were calculated using Student's t-test. Logistic regression and decision tree were built for cancer detection and their performances were evaluated by the area under the curve (AUC). Twenty-four genomic locations were differentially methylated, which could be mapped to eleven genes. Nine out of eleven genes had differentially expressed mRNA levels, which were used to build the model for cancer detection. The final detection models consisting of mRNA expression levels of these nine genes had great performances on both training group and testing group. The model that constructed in this study suggested MSX1 and DCLK1 might be used in colorectal cancer detection or as target of cancer therapies. J. Cell. Physiol. 232: 1879-1884, 2017. © 2016 Wiley Periodicals, Inc.
© 2016 Wiley Periodicals, Inc.