A Hierarchical Hidden Markov Model for the Annotation of Chromatin States

• Activity Number: 594
• Type: Topic Contributed
• Date/Time: Wednesday, August 3, 2016 : 2:00 PM to 3:50 PM
• Sponsor: International Chinese Statistical Association
• Author(s): Guo-Cheng Yuan* and Eugenio Marco and Wouter Meuleman and Manolis Kellis and Jialiang Huang and Kimblerly Glass and Jianrong Wang and Luca Pinello
• Companies: Dana-Farber Cancer Institute and Editas Medicine and MIT and MIT and DFCI and Harvard Medical School and MIT and DFCI
• Keywords: chromatin ; hidden Markov model ; epigenetics ; ENCODE ; domain

Abstract

Genome-wide annotation of chromatin states through mapping of chromatin regulators and histone marks has led to important insights in gene regulation. However, existing methods are limited in that they cannot resolve patterns at multiple length scales. We present diHMM, a novel computational method that systematically characterizes chromatin states at multiple levels. We applied diHMM to analyze a ChIP-seq dataset containing 9 chromatin marks in the three Tier 1 cell lines from ENCODE. We found that the domain-level states correctly recapitulated the large-scale domains; at the same time, the nucleosome-level state organization preserves the high-resolution information detected from ChromHMM. Our analysis suggests that the usage, dynamics, and biological function of a nucleosome-level state are highly dependent on the domain-level context. Our method is generally applicable to gene regulatory analysis in various biological systems.