Department of Biostatistics and Computational Biology
Dana-Farber Cancer Institute
Hierarchical Annotation of Chromatin States in Stem Cells
Epigenetic mechanisms play an important role in many diseases, but our mechanistic understanding of epigenetic regulation is still incomplete. One major difficulty is that chromatin forms complex three-dimensional structures and studies have failed to map genome-wide chromatin interactions with enough resolution. On the other hand, genome-wide distributions of the first-order chromatin structure, such as histone modifications, have been characterized at increasingly higher resolution. In order to identify multi-layer chromatin structures simultaneously, we have developed a Hierarchical Hidden Markov Model (HHMM) with two-layers of chromatin states, which we call domain- and bin-level states, respectively. Using this method, we analyzed a ChIPseq dataset of 9 histone marks in human embryonic stem cells, and identified a number of chromatin domains that can be validated by independent studies. At the same time, bin-level states detected variations in histone modification patterns at high resolution. Our new HHMM approach has uncovered higher order chromatin states and provides novel insights into epigenetic regulation in normal development and disease.