Integrated flexible DNA methylation-chromatin segmentation modeling enhances epigenomic state annotation

Abstract

DNA methylation and histone modifications together shape the cell-type-specific epigenomic landscape. To enhance genome-wide annotation, we developed EpiSegMixMeth (ESMM), the first integrative segmentation model combining chromatin marks and DNA methylation. ESMM improves upon hidden Markov models by incorporating flexible read count distributions and state duration modeling. Applied to 154 high-quality human epigenomes from the IHEC EpiATLAS, ESMM enhances the annotation of broad heterochromatic regions—over 60% of the genome— that are often missed by chromatin-only models. It accurately defines narrow regulatory element boundaries and captures local chromatin state transitions during cell differentiation. Notably, we show that DNA methylation can replace missing repressive histone marks in segmentation, ensuring robust results across various cell types. In developing memory B cells, ESMM reveals chromatin shifts that align with 3D genome architecture changes, providing a valuable resource for studying cell-type-specific epigenomic regulation.