has anyone some experience with the use of ChromHMM for chromatin segmentation and can explain me how to get annotations for the generated chromatin states?
I applied ChromHMM, that means the functions 'binarizeBed' and 'LearnModel', to a sample with four marks (H3K27ac, H3K27me3, H3K4me1 and H3K4me3). Is set the number of chromatin states to 6... just to try if the commands run with my input files. Now, I have two questions/problems:
- How many chromatin states can I predict on the basis of 'only' four chromatin marks? I assume that it makes no sense to generate 15 chromatin states or more based on my limited data set, or am I wrong?
- My more urgent problem: How can I annotate my chromatin states with "active Promotor", "weak Enhancer" etc? In the LearnModel output, the six states are simply numbered consecutively (E1-E6). Does somebody know if ChromHMM provides a solution for annotation? In fact, at Roadmap there are stored labelmap files with annotations for the 15, 18 and 25 state models which can be considered in the command 'MakeBrowserFiles' by the parameter '-m labelmappingfile' but I don´t know if/how I can use these files for my segmentation...
Thank you very much in advance for your comments and help!