I actually try to run ChromHMM on a new cell line for which I have the following markers (H3K4me1, H3K4me3, H3K9me3, H3K27ac, H3K27me3, H3K36me3, Dnase Hotspot and input) and I have some difficulties to retrieve right annotation corresponding to different chromatin state. Does somebody have some advices/tips to share?
Does somebody already try this software on a new cell line?
Can you clarify "retrieve the right annotation"? Are you asking about where to get more data, or are you asking about how to interpret the ChromHMM output?
Sorry for the misunderstood, i was talking about how to interpret the ChromHMM output and more specially how do you assess the right annotation for each state.
It's an art more than a science. It takes a bit of staring at the emission probabilities. Use the OverlapEnrichment tool with a bunch of BED files of factors whose biological function is well-studied to help interpret the models. For example in your case high H3K4me1, low H3K4me3, and high H3K27ac for some state N suggests a label of "enhancer-like" for state N. You could use OverlapEnrichment with, say, P300 peaks (something not used in the model) to help justify that label. Enhancers, active promoters, and repressive chromatin are usually pretty straightforward to label. The other states that don't have an obvious succinct label get tricky and end up with awkward names. That's biology for you though -- defying simplicity!
Can you clarify "retrieve the right annotation"? Are you asking about where to get more data, or are you asking about how to interpret the ChromHMM output?
Hi,
Sorry for the misunderstood, i was talking about how to interpret the ChromHMM output and more specially how do you assess the right annotation for each state.