Classification Based on Feature Extraction for Hepatocellular Carcinoma Diagnosis Using High-throughput Dna Methylation Sequencing Data

DNA methylation is a well-studied mechanism of epigenetic regulation, which plays an important role in oncogenesis and tumor progression. Even at very early stage, cancer genome exhibits aberrant methylation patterns, such as hypermethylation and hypomethylation at different scales. The detection of abnormal methylation patterns with whole-genome bisulfite sequencing (WGBS) using circulating DNA from plasma has become a promising method for cancer diagnosis. In this study, Boruta, an extension of the random forest, was used to select important features (variables). Those selected features were used to establish a support vector machine (SVM) classifier for liver cancer diagnosis. As the results, a WGBS data set from hepatocellular carcinoma (HCC) patients was employed to show the improved performance of the proposed method for diagnosis.