The directories contain summary ChromHMM annotations for a group of samples (denoted a *combination*) for the 18-state per sample annotations. 
The states' names, biological characterizations are similar to those generated by Roadmap's 18-state annotation 
(https://egg2.wustl.edu/roadmap/web_portal/chr_state_learning.html#exp_18state)

Summary ChromHMM tracks were generated for each combination of samples from IHEC consortium. 
The combinations are grouped based on columns: harmonized_sample_ontology_intermediate and 
harmonized_sample_disease_high in the metadata. We only calculated the summary chromatin state maps for combinations with at least *2* samples. 
In total, there are 56 combinations. 
 
The combination name was obtained by appending to harmonized_sample_ontology_intermediate the suffix H, D, or C corresponding to
to healthy, disease, or cancer entries in harmonized_sample_disease_high

For the 50 combinations with fewer than 100 samples summary annotations were generated with the CSREP method, as described in
Vu H, Koch Z, Fiziev P, Ernst J.
A framework for group-wise summarization and comparison of chromatin state annotations.
Bioinformatics, 39:btac722, 2023.

For the 6 combinations with >=100 samples, summary annotations were generated by simply taking the fractions, at each genomic positions, of total samples 
that are annotated as each state, i.e. if 3 out of 5 samples are annotated as state 1_TssA at position i, then this position is summarized as 
state 1_TssA with probability 0.6. The six such combinations were Brain Disease (brainD), Brain Healthy (brainH), Monocyte healthy (monocyteH), 
Neutrophil healthy (neutrophilH), t cell healthy (t_cellH), and venous blood cancer (venous_bloodC).

The files are arranged in three subdirectories
SUMMARYCHROMHMM_BED - contains a four column bed file with the summary ChromHMM annotation for each combination. Files are named in the format <combination_name>_summary_18_ChromHMM.bed.gz . Example: colonC_summary_18_ChromHMM.bed.gz (note combination_name can have underscores corresponding to spaces, ex: erythroid_ageH_summary_18_ChromHMM.bed.gz)
SUMMARYCHROMHMM_BIGBED - contains a bigbed file with the summary ChromHMM annotation for each combination. Files are named in the format <combination_name>_summary_18_ChromHMM.bb. Example: colonC_summary_18_ChromHMM.bb (note combination_name can have underscores corresponding to spaces. Example: erythroid_ageH_summary_18_ChromHMM.bb)
SUMMARYCHROMHMM_PROBABILITIES_BIGWIG - contains a bigwig with the state estimate probabilities for each combination and state pair. Files are named in the format <combnation_name>_<statenumber_statemnemonic>.bw. Example: kidneyC_8_EnhG2.bw

The statenumber is 1-based (1-->18). The state mnemonic is similar to those in 18-state model from Roadmap (link: https://egg2.wustl.edu/roadmap/web_portal/chr_state_learning.html#exp_18state)

All data here are generated in hg38.

Within each sub-directory there is a .md5 with the checksum for all the files in the subdirectory.

The tracks were generated by Ha Vu.
For correspondence contact Jason Ernst (jason.ernst@ucla.edu)