This document summarizes key findings from an analysis of 111 reference human epigenomes. It finds that 1) histone mark combinations predict gene expression and have distinct methylation and accessibility profiles, 2) megabase domains show differences in activity and structure, and 3) enhancers are enriched for conserved elements and coordinated modules associated with phenotypes.

1. Reading circle of Epigenome Roadmap
doi:10.1038/nature14248
Roadmap Epigenomics Consortium et. al. Integrative
analysis of 111 reference human epigenomes
Itoshi NIKAIDO, Ph.D. <itoshi.nikaido@riken.jp>
Unit Leader, Bioinformatics Research Unit
RIKEN Advanced Center for Computer and Communication
http://bit.accc.riken.jp/

2. Main figure: 9
Extended Data Figure: 12
Supplementary Figure: 13

3. 1. Reference epigenome mapping across tissues and cell types
2. Chromatin states, DNA methylation and DNA accessibility
3. Epigenomic differences during lineage specification
4. Most variable states and distinct chromosomal domains
5. Relationships between marks and lineages
6. Imputation and completion of epigenomic data sets
7. Enhancer modules and their putative regulators
8. Impact of DNA sequence and genetic variation
9. Trait-associated variants enrich in tissue-specific marks
10. Discussion
Headlines
Integrative analysis of 111 reference human epigenomes

4. 1. Histone mark combinations show distinct levels of DNA methylation and accessibility, and
predict differences in RNA expression levels that are not reflected in either accessibility or
methylation.
2. Megabase-scale regions with distinct epigenomic signatures show strong differences in activity,
gene density and nuclear lamina associations, suggesting distinct chromosomal domains.
3. Approximately 5% of each reference epigenome shows enhancer and promoter signatures,
which are twofold enriched for evolutionarily conserved non-exonic elements on average.
4. Epigenomic data sets can be imputed at high resolution from existing data, completing missing
marks in additional cell types, and providing a more robust signal even for observed data sets.
5. Dynamics of epigenomic marks in their relevant chromatin states allow a data-driven approach
to learn biologically meaningful relationships between cell types, tissues and lineages.
6. Enhancers with coordinated activity patterns across tissues are enriched for common gene
functions and human phenotypes, suggesting that they represent coordinately regulated modules.
7. Regulatory motifs are enriched in tissue-specific enhancers, enhancer modules and DNA
accessibility footprints, providing an important resource for gene-regulatory studies.
8. Genetic variants associated with diverse traits show epigenomic enrichments in trait-relevant
tissues, providing an important resource for understanding the molecular basis of human disease.
Summary: 8 findings
Integrative analysis of 111 reference human epigenomes

5. Pegs and guy-rope modelFigure 1
Tissues and cell types profiled in the Roadmap Epigenomics

6. 127 reference epigenome
= 111 epigenome roadmap + 26 ENCODE
!
2,805 genome-wide data sets =
1,821 histone modification data sets
360 DNA accessibility data sets
277 DNA methylation data sets
166 RNA-seq
!
Chromatin immunoprecipitation (ChIP)
DNA digestion by DNase I (DNase)
Bisulfite treatment
Methylated DNA immunoprecipitation (MeDIP)
Methylation-sensitive restriction enzyme digestion (MRE)
RNA profiling

7. Figure 2
Data sets available for each reference epigenome.

8. What is chomatin states?
Prediction of 51 chromatin states from epigenome data set
• 51 chromatin states
•11 Promoter states
•high/mid/low expression, repressed, high/mid/low GC, …
• 17 Transcribed States
•5’ proximal high/mid expression, open chromatin, TF
binding, spliced exon,…
• 11 Active intergenic states
• strong/weaker/distal/proximal enhancer, CTCF, H2AZ,…
• 6 Repressed states
• unmappble, A/T rich, ERVL, heterochromatin…
• 6 Repetitive states
• (CA)n, (TG)n, L1/LTR, Satellite repeats
http://www.nature.com/nbt/journal/v28/n8/extref/nbt.1662-S1.pdf

9. What is chomatin states?
Prediction of 51 chromatin states from epigenome data set
Epigenome Marks (observed)
Chromatin States (Predicted)
http://www.nature.com/nbt/journal/v28/n8/fig_tab/nbt.1662_F1.html
Circles are 200 bp windows
of chromatin c.
Pt-1 Pt Pt+1
Vt-1 Vt-1 Vt+1

10. What is chomatin states?
ChromHMM: automating chromatin-state discovery and characterization
http://www.nature.com/nbt/journal/v28/n8/extref/nbt.1662-S1.pdf
http://www.nature.com/nbt/journal/v28/n8/full/nbt.1662.html#supplementary-information

11. Figure 3
Epigenomic information across tissues and marks
Red: constitutive promoter
Yellow: active enhancer

12. Figure 4a-e
Chromatin states and DNA methylation dynamics
15 states model: 5 histone modification, 127 epigenomes
1. High expression = low methylation + high accessibility
2. Low expression = high methylation + low accessibility
3. Enhancer = intermediate methylation + intermediate accessibility

13. Figure 4a-e
Chromatin states and DNA methylation dynamics
• TxFlnk, Enh, TssBiv and BivFlnk states show similar distributions of
DNA accessibility but different distributions of gene expression and
DNA methylation.
• Enh and PeprPC states show similar distributions of DNA
methylation but different distributions of DNA accessibility
• etc…
Complex relationship:
active or repressive region << chromatin states

14. Figure 4g
Chromatin states and DNA methylation dynamics
Methylation, 95 epigenomes
• TssAFlnk: unmethylated in differentiated cells and tissues
• Enh: Highly methylated in ES and iPS
• EnhBiv: broad distribution in ES and iPS (cell-to-cell variability)
• PeprPC: varying methylation levels among cell and tissues
DNA methylation changes
during ESC differentiation

15. Figure 5ab
Cell-type differences in chromatin states
constitutive
Cell specific
• HSC: ↑TxWk, ↓TssA/TssBiv
• ESC: ↑TssBiv, ↓PeprPCWk (restriction of
H2K27me3-establishing Polycomb at promoter)
• IMR90: ↑Het/ReprPC/EhnG, ↓Quies
Variability of chromatin states Chromatin state frequency

16. Figure 5cd
Cell-type differences in chromatin states
Transition of chromatin state Chromatin states at a larger resolution (2Mb)

17. Figure 6
Epigenome relationships

18. Figure 7
Regulatory modules from epigenome dynamics

19. Figure 8
Linking regulators to target tissues and cell types

20. Figure 9
Linking regulators to target tissues and cell types

21. 1. Reference epigenome mapping across tissues and cell types
2. Chromatin states, DNA methylation and DNA accessibility
3. Epigenomic differences during lineage specification
4. Most variable states and distinct chromosomal domains
5. Relationships between marks and lineages
6. Imputation and completion of epigenomic data sets
7. Enhancer modules and their putative regulators
8. Impact of DNA sequence and genetic variation
9. Trait-associated variants enrich in tissue-specific marks
10. Discussion
Headlines
Integrative analysis of 111 reference human epigenomes

22. 1. Histone mark combinations show distinct levels of DNA methylation and accessibility, and
predict differences in RNA expression levels that are not reflected in either accessibility or
methylation.
2. Megabase-scale regions with distinct epigenomic signatures show strong differences in activity,
gene density and nuclear lamina associations, suggesting distinct chromosomal domains.
3. Approximately 5% of each reference epigenome shows enhancer and promoter signatures,
which are twofold enriched for evolutionarily conserved non-exonic elements on average.
4. Epigenomic data sets can be imputed at high resolution from existing data, completing missing
marks in additional cell types, and providing a more robust signal even for observed data sets.
5. Dynamics of epigenomic marks in their relevant chromatin states allow a data-driven approach
to learn biologically meaningful relationships between cell types, tissues and lineages.
6. Enhancers with coordinated activity patterns across tissues are enriched for common gene
functions and human phenotypes, suggesting that they represent coordinately regulated modules.
7. Regulatory motifs are enriched in tissue-specific enhancers, enhancer modules and DNA
accessibility footprints, providing an important resource for gene-regulatory studies.
8. Genetic variants associated with diverse traits show epigenomic enrichments in trait-relevant
tissues, providing an important resource for understanding the molecular basis of human disease.
Summary: 8 findings
Integrative analysis of 111 reference human epigenomes