The following instructions are for use in terminal of Unix (or Mac) machine. 

They require BEDtools and bash shell to be installed. 

Use "tar -xvf <archive>" command to decompress the archived folders. 

1. Binarization of the signal (only necessary to do if you want to add or remove marks to ones already present)

Generates binary ChromHMM input files from SICER BED files using the following operations:

A) Use the following command to make 200bp windows of the whole genome:

bedtools makewindows -g mm9.chromsizes -w 200 > 200.mm9_windows.bed

B) Move translate.sh, AGR_narrow.bed, and 200.mm9_windows.bed files to the folder with SICER BED files. Then run

./translate.sh <cell_name> 

where <cell_name> is cell type name, in our case "proB". 

Notice that the following naming convention  for the BED files should be observed: <cell_name>_<feature_name>.bed

As a result of this script, you will obtain 21 file ending in .txt, one for each chromosome. 

Place them in a separate directory and use it as input for ChromHMM. 

2. Running ChromHMM

Download ChromHMM from http://compbio.mit.edu/ChromHMM/ and read the manual. 

In order to run ChromHMM in the terminal, use the following command: 

java -mx8000M -jar ChromHMM.jar LearnModel <input_dir> <output_dir> <number_of_states> mm9

where 
  - <input_dir> is the directory with all 21 .txt files (see above); 
  - <output_dir> is the name of the directory where you want your output; 
  - <number_of_states> is (mostly arbitrary) number of states for the Hidden Markov Model. 

For example, the unarchived "proB_input" directory could be used as input.