Charting differentially methylated regions in cancer with Rocker-meth

Differentially DNA methylated regions (DMRs) inform on the role of epigenetic changes in cancer. We present Rocker-meth, a new computational method exploiting a heterogeneous hidden Markov model to detect DMRs across multiple experimental platforms. Through an extensive comparative study, we first demonstrate Rocker-meth excellent performance on synthetic data. Its application to more than 6,000 methylation profiles across 14 tumor types provides a comprehensive catalog of tumor type-specific and shared DMRs, and agnostically identifies cancer-related partially methylated domains (PMD). In depth integrative analysis including orthogonal omics shows the enhanced ability of Rocker-meth in recapitulating known associations, further uncovering the pan-cancer relationship between DNA hypermethylation and transcription factor deregulation depending on the baseline chromatin state. Finally, we demonstrate the utility of the catalog for the study of colorectal cancer single-cell DNA-methylation data. Matteo Benelli et al. present Rocker-meth, a new Hidden Markov Model (HMM)-based method, to robustly identify differentially methylated regions (DMRs). They use Rocker-meth to analyse more than 6000 methylation profiles across 14 cancer types, providing a catalog of tumor-specific and shared DMRs.

Automated summary

Researchers present a new computational method, Rocker-meth, which utilizes a heterogeneous hidden Markov model to detect differentially methylated regions (DMRs) in cancer. Through analysis of over 6,000 methylation profiles across 14 cancer types, Rocker-meth efficiently identifies tumor-specific and shared DMRs.