Regions delineated using observed DNaseI data across 53 epigenomes, annotated with the 5-mark 15-state model based on observed data across 127 epigenomes (Roadmap + ENCODE).
All coordinates are based on the Human hg19 genome assembly.
Visualization of these putative regulatory regions in the context of the chromatin states can be seen here.
Please indicate whether you're using any of these data by filling out this spreadsheet.

DNaseI regions selected with -log10(p) >= 2

All DNaseI region coordinates, regardless of region type (prom, enh, dyadic, rest) in one BED file. The score column contains a rough indication of DNaseI signal strength in the epigenomes which had a DNaseI peak at that position.

Promoters: 84,880 putative promoter regions (1.55% of genome)

ChromHMM state calls (RData)
BED files per epigenome
Clustered with k=85, for a resolution of ~1k regions per cluster

Enhancers: 2,464,858 putative enhancer regions (14.29% of genome)

ChromHMM state calls (RData)
BED files per epigenome
Clustered with k=246, for a resolution of ~10k regions per cluster

Dyadics: 209,812 putative dyadic regions (1.64% of genome)

ChromHMM state calls (RData)
BED files per epigenome
Clustered with k=210, for a resolution of ~1k regions per cluster

DNaseI regions selected with -log10(p) >= 10

All DNaseI region coordinates, regardless of region type (prom, enh, dyadic, rest) in one BED file. The score column contains a rough indication of DNaseI signal strength in the epigenomes which had a DNaseI peak at that position.

Promoters: 45,501 putative promoter regions (1.24% of genome)

ChromHMM state calls (RData)
BED files per epigenome
Clustered with k=46, for a resolution of ~1k regions per cluster

Enhancers: 618,429 putative enhancer regions (7.46% of genome)

ChromHMM state calls (RData)
BED files per epigenome
Clustered with k=62, for a resolution of ~10k regions per cluster

Dyadics: 54,269 putative dyadic regions (0.75% of genome)

ChromHMM state calls (RData)
BED files per epigenome
Clustered with k=54, for a resolution of ~1k regions per cluster

For questions/comments, contact Wouter Meuleman (lastname at mit edu)