Prognostic biomarker identification and tumor classification in breast cancer patients by methylation and transcriptome analysis

Abstract

Breast cancer is one of the most common and heterogeneous malignancies. Although the prognosis of breast cancer has improved with the development of early screening, the mechanisms underlying tumorigenesis and progression remain incompletely understood. DNA methylation has been implicated in tumorigenesis and tumor development and, in the present study. we screened methylation-driven genes and explored their prognostic values in breast cancer. RNA-sequencing (RNA-Seq) transcriptome data and DNA methylation data of the TCGA-BRCA dataset were obtained from The Cancer Genome Atlas. Differentially expressed genes and differentially methylated genes were identified separately. The intersected 783 samples with both RNA-Seq data and DNA methylation data were selected for further analysis. Fifty-six methylation-driven genes were identified using the MethylMix r package and 10 prognosis methylation-driven genes (CDO1, CELF2, ITPAIPL1, KCNH8, PTK6, RAB25, RIC3, USP44, ZSCAN1 and ZSCAN23) were further screened by combined methylation and gene expression analysis. Based on the methylation data of the screened 10 methylation-driven genes, six subgroups were identified with the ConsensusClusterPlus r package. The protein levels of the 10 prognostic methylation-driven genes were detected by immunohistochemical experiments. Moreover, based on the RNA-Seq data, a signature calculating the risk score of each patient was developed with stepwise regression. The risk score and other clinical features (age and stage) were confirmed to be independent prognostic factors by univariate and multivariate Cox regression analyses. Finally, a prognostic nomogram incorporating all the significant factors was integrated to predict the 3-, 5- and 7-year overall survival. Taken together, the methylation-driven genes identified here may be potential biomarkers of breast cancer.