Tools to identify enhancers from chromatin marks

Question

[guiu]

Hello everyone,

I would like to know if there are tools that feeded with ChIP-seq data for various histone marks, identify the enhancers by the "right" methylation/acetylation combinations. Until now, I found CSI-ANN, but I was wondering if there are other tools.

Thanks

Answer 1

[Ryan Dale]

The search term you're looking for is "chromatin segmentation". For example, omicstools has a list of chromatin segmentation tools

Of these, ChromHMM is probably the most widely used in publications at the moment.

Also see other answers on biostars on chromatin segmentation and enhancers:

• ChromHMM annotation new cell line
• Predictability Of Regulatory States In New Cell Lines Or Tissues From Existing Encode Segmentations?
• Meaning of Chromatin state
• Chromatin state segmentation - direction of effect of the identified enhancer

[Sinji]

I'll supplement Ryan's excellent answer with more information on the chromatin signature of Enhancers.

Generally as peeyushashu stated, Enhancers are denoted by a high amount of H3K4me1. This seems to be the single agreed on feature in published papers.

Outside of that you have varying 'classes' of enhancers. The three most common are those termed Active, Primed, and Latent Enhancers. Active enhancers are typically DNase hypersensitive sites, with H3K27Ac methylation, Primed enhancers are marked with H3K4me1, low levels of H3K27Ac and are DNase hypersensitive, while latent enhancers show low or absent H3K4me1, H3K27Ac and are often not DNase hypersensitive which means that chromatin at that region is closed.

I'd also like to emphasize that Enhancers do not have to be absent of H3K4me3. There's literature that has shown that Enhancers marked with H3K4me3 are considered to be active.

There are other enhancer classes that are denoted by H3K27me3, H2A.Z, and a few other markers I can't think off the top of my head, so make sure you read around.

[guiu]

Yes I have been reading aroung, actually I am looking for a tool because I know that there are multiple signatures that identify enhancers, so I am afraid of missing out some combination while overlapping the different chromatin marks myself. I thought a tool intended for this would check for all the combinations (given I provide the data for different histone marks of course). Thanks for the overview anyway, much appreciated.

[guiu]

Thanks so much Ryan, finding the search terms is always the most difficult to me. I will definitely try ChromHMM.

[Sinji]

It's useful to keep in mind that ChromHMM does not annotate the different enhancer combinations for you. You'll have to do the annotation yourself.

[Ryan Dale]

Yep -- see specifically A: ChromHMM annotation new cell line

[guiu]

I read something about ChromHMM, but I have a doubt. Why would I use it and "learn" from my data, instead of just looking at the overlap of different histon marks? Especially in the case when I am not searching for something "new", I just want to identify the active enhancers in my specific cell line. Does it make sense to complicate things with machine learning in this case?

[Ryan Dale]

The problem is that there is not a good definition of "enhancer". Do you use "overlap of H3K4me1 and H3K4me3"? Or maybe "high ratio of H3K4me1 to H3K4me3". Maybe require DNase hypersensitivity. And maybe require some threshold of acetylation. Ideally you would also require evidence of long-range interaction with at least one promoter. There are many differing definitions in the literature along these lines. Another option is to use chromatin segmentation tools which find patterns in your data that you can then interpret as "enhancer-like". An advantage to this is that you can find some interesting results, like different classes of enhancers that you otherwise might not have found. And you can always use the ad hoc enhancer definitions to look for enrichment in chromatin segmentations.

Answer 2

[SP]

I have not used any tools but an easy thing could be to do a overlap between e.g. H3K27ac and H3K4me1 and found the overlapping regions, these are the most likely enhancer marks. You can do a heatmap for H3K27ac peaks and compare the occurrence of H3K4me1 and H3K4me3 around these peaks. you can use seqMiner for producing a heatmap. Regions which are enriched with K27ac + K4me1 and Edited: absent or low K4me3 are most probable enhancer regions (if you have region annotation of these peaks, these will mostly be intergenic of intronic peaks). You need to call for H3K27ac peaks before the analysis.

[guiu]

Thanks a lot. Just a small follow-up, so is it only overlap what counts? I thought also the height/shape of the peaks would be used to make the identification of enhancers better.

[SP]

I am not sure how much peaks shape and hight will help, every histone modification has it's own pattern. Of-course there are difference in shape or intensity of peaks on enhancer and promoter but for me combination of histone marks is a better proof. I would like to know what fellow-researchers think about this?