A tiered hidden Markov model characterizes multi-scale chromatin states

Genomics. 2013 Jul;102(1):1-7. doi: 10.1016/j.ygeno.2013.03.009. Epub 2013 Apr 6.


Precise characterization of chromatin states is an important but difficult task for understanding the regulatory role of chromatin. A number of computational methods have been developed with varying levels of success. However, a remaining challenge is to model epigenomic patterns over multi-scales, as each histone mark is distributed with its own characteristic length scale. We developed a tiered hidden Markov model and applied it to analyze a ChIP-seq dataset in human embryonic stem cells. We identified a two-tier structure containing 15 distinct bin-level chromatin states grouped into three domain-level states. Whereas the bin-level states capture the local variation of histone marks, the domain-level states detect large-scale variations. Compared to bin-level states, the domain-level states are more robust and coherent. We also found active regions in intergenic regions that upon closer examination were expressed non-coding RNAs and pseudogenes. These results provide insights into an additional layer of complexity in chromatin organization.

Publication types

  • Research Support, N.I.H., Extramural
  • Research Support, U.S. Gov't, Non-P.H.S.

MeSH terms

  • Binding Sites
  • Chromatin / genetics*
  • Chromatin / ultrastructure
  • Computational Biology*
  • DNA / chemistry
  • DNA, Intergenic / chemistry
  • DNA, Intergenic / ultrastructure
  • Embryonic Stem Cells / chemistry
  • Embryonic Stem Cells / cytology*
  • Epigenesis, Genetic
  • Histones / chemistry
  • Humans
  • Markov Chains
  • RNA, Untranslated / genetics*


  • Chromatin
  • DNA, Intergenic
  • Histones
  • RNA, Untranslated
  • DNA