Screening Dys-Methylation Genes and Rules for Cancer Diagnosis by Using the Pan-Cancer Study

• Published in: IEEE access 2020, Vol.8, p.489-501
• Main Authors: Zhang, Yu-Hang, Zeng, Tao, Pan, Xiaoyong, Guo, Wei, Gan, Zijun, Zhang, Yunhua, Huang, Tao, Cai, Yu-Dong
• Format: Article
• Language: English
• Subjects: Bioinformatics; Cancer; cancer diagnosis; Decision trees; Deoxyribonucleic acid; Diagnosis; Diagnostic systems; DNA; DNA methylation; Genomics; Medical diagnosis; Methylation; pan-cancer; Public health; Radio frequency; rule; Signs and symptoms; Tumors
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2019.2961402

Although cancer has long been a major public health problem worldwide, conquering cancer still remains unachievable due to its complexity and diversity. With the development of high-throughput sequencing technologies, the combination of conventional clinical symptoms and new genetic events is becoming effective and has been approved for precise prediction and innovative diagnostic strategies. Epigenetic modification (e.g., DNA methylation) is an important mechanism of transcriptional control in normal or disease functions. Depending on the unique methylation profiles, an important possibility raised is that the characteristic of dys-methylation could provide a new molecular marker system for the identification of the major forms of tumors. In this study, we attempted to distinguish different tumor types. On the basis of DNA methylation data from PanCanAtlas in The Cancer Genome Atlas, we applied mRMR and MCFS methods together to identify the decision rules for distinguishing 33 different tumor types and ranked the features that characterized methylation level. This study highlights the considerable application potential of methylation features in cancer diagnosis and provides insight into novel therapeutic targets.