Elucidating Cancer Subtypes by Using Epigenome and Genome Cross-Talk

DNA methylation plays a critical role in tumorigenesis and tumor malignancy . The role of this epigenetic phenomena, especially, in the investigation of cancer subtypes remains rather under-explored. In this study, we use sCClust (sparse Canonical Correlation analysis with Clustering), a method that combines high-dimensional omics data using sparse canonical correlation analysis (sCCA), to gene expression and DNA methylation data for prostate cancer to elucidate underlying subtypes. We compare the subtypes with the TCGA taxonomy and evaluate the results using survival analysis. Our findings demonstrate that data integration results in statistically significant subtypes similar to the TCGA subtypes and outperforms existing classification efforts. The significance of this study lies in enhancement in the subtyping of prostate cancer, as well as statistical integration of epigenomic and transcriptomic data.