Motif analysis from H3K27ac ChIPseq broad peaks

Question

[Prakash]

Hi all.

Recently, I got sequence for H3K27ac ChIP-seq data performed in various stimulation and knockdown condition. I have processed the reads and did peak calling using macs2. Looking into IGV browser, the enrichment looks good and the peaks are identified. The peaks called are of in the size range from ~1kb to 10kb. My concern here is whether the size of peaks identifies is really broad or is it ok to get large size peaks from H3k27ac chip. Next if want to perform motif identification, should I take all the broad peaks or perform peak calling to get broad peaks of lesser size (~1kb).

I would highly appreciate your suggestions and feedbacks

Thanks

peak calling method:

ls bowtie_out/*.25M.bam | cut -d'/' -f2 | parallel --verbose 'macs2 callpeak -t bowtie_out/{} -c ../Input_R1.bam -g mm -n {=s/.25M.bam//g=} --verbose 2 --broad --broad-cutoff 0.01 --cutoff-analysis  --fe-cutoff 4 --outdir macs_peak'

IGV browser snapshot:

Zoom of portion highlighted in above image:
free image hosting

[gable_works]

The motifs bound by TFs are most likely within the valleys of your H3K27ac signal. This hypothesis has been tested (article) and a program called EpiSAFARI was developed to identify peaks within valleys . You can certainly do motif analysis on the peak set you have but it wouldn't make must biological sense. I also want to point out that this is true for H3K27ac but may not be true for other histone modifications.

Answer 1

[venu]

If the signals in IGV are looking reasonable over background for ~10kb domains, it is totally acceptable. There are something called Super-enhancers which are minimum of 10-12kb in size with high H3K27ac signal. Just google the term, you'll find number of papers solely based on super-enhancers.

Regarding motif analysis, instead of using total peak length, identify nucleosome free region within your H3K27ac peaks. These are potentially identified by small valleys which lack signal within peaks. Homer tool has a function called nfr which is what you want.

BTW, is there a specific reason to use broad option with MACS in your case? In my experience, narrowpeaks from MACS are of good quality and works very well for all types of analyses.

[Prakash]

Thanks @Venu for your insight. This was helpful. The reason using broad option is that histone acetylation marks covers broad regions (~1kb). I tried calling both using narrow and broad options, in my case also the narrow peaks seems working well. As per your suggestion i can use narrow peaks and identify those nfr for motif analysis.

upload