ChromHMM results from Jason Ernst Lab (jason.ernst@ucla.edu)
=======================================================

Per-epigenome annotations were generated by applying the ChromHMM (Ernst and Kellis, 2012) (v1.24) MakeSegmentation command with the 18-state model from Roadmap Epigenomics Consortium et al, Nature 2015. The 18-state model is based on H3K4me1, H3K4me3, H3K9ac, H3K27ac, H3K27me3, H3K36me3. In total annotations were generated for 1698 epigenomes. Only chr1-22 and chrX were annotated.

Observed data was used when available and otherwise used imputed data from ChromImpute (Ernst and Kellis, 2015).
Observed data was binarized relative to its control file with the BinarizeBam command.
The '-paired' flag was included for those datasets annotated as being based on paired end reads.
The default resolution (200bp) and binarization threshold of BinarizeBam was used to binarize the observed data.

For imputed data, the data was converted from 25bp to 200bp resolution by averaging the imputed values in each 200bp bin. The binarization threshold was set in a mark specific way so that the average fraction of bins receiving a present call overall datasets for imputed data for the mark matched that average fraction for all observed data.

./BED/ folder contains ChromHMM annotations in the bed format; all md5 sums are included.
./bigBed folder contains ChromHMM annotations for viewing in a browser in bigBed format; all md5 sums are included.
./model_18_core_K27ac.txt ChromHMM model file for the 18-state model
./color_model_18_core_K27ac.txt file mapping 18-states to R,G,B colors
./label_model_18_core_K27ac.txt file mapping 18-states to state labels
./ChromHMM_per_chr_matrix// contains summary per chromosome matrices with 200bp resolution (1st column - to get genomic position 1st column has to be multiplied by 200) - 
containing numerical 'ChromHMM state index' for all epigenomes. All matrices are in ',' separated 'csv' format. (These were produced by Misha Bilenky of BCGSC)
File ./ChromHMM_per_chr_matrix//header contains header fields (as column) common for all matrices; 1st field is a start position of the bin on chromosome; rest are epigenome labels.
File ./ChromHMM_per_chr_matrix//states.txt - ChromHMM states - index correspondence