I'm analyzing Chip-Seq data and am interested in histone modifications that have very broad peaks such as H3K27me3 and H3K36me3. I have tried peak calling using Homer but it does not seem to work well. Is there a better peak caller for these modifications that have broad enrichment across the genome? Thanks
I have some collections here https://github.com/crazyhottommy/ChIP-seq-analysis#peak-calling
SICER is a specialized peak caller for broad enriched histone modification (Zang, et al 2009). A comment in OMICtools also said that SICER is better than MACS for broad enrichment antibodies. However, I use MACS 1.4, the most stable version, to call H3K27me3 peaks with parameters suggested by Dr. Liu as below (Feng, et al 2012). For our ChIP-Seq data, MACS 1.4 is good enough. I used MACS, but not SICER, because we would like to use the same peak calling tool to call all ChIP-Seq data, including broad and sharp enriched peaks. In addition, MACS 2 has added broad enriched parameters (--broad) that could be helpful for your purpose.
Ideal for broad peaks like H3K27me3 and H3K36me3. Where MACS will truncate peaks in smaller fragments, generating dozens of small peaks, the strenght of SICER is that you can take gaps into account. You just have to perform various tests to choose a good combination between gap size and window size. Typically, I used a window of 500nt and a gap of 3500nt for a good compromise between sensitivity and specificity.
Posted by Fabien Pichon on Oct 17, 2014
> macs14 -t BROAD_GM12878_H3K4me3.bam –c BROAD_GM12878_H3K4me3_Control.bam -g hs -n BROAD_GM12878_H3K4me3 --nomodel --shiftsize 73 -B -S --call-subpeaks
I am biased, but https://github.com/endrebak/epic is good.
Also, if you prefer R and know some linear modelling csaw is an option: https://bioconductor.org/packages/release/bioc/html/csaw.html
You could try ZINBA http://genomebiology.com/2011/12/7/r67 "ZINBA integrates local covariates with DNA-seq data to identify broad and narrow regions of enrichment, even within amplified genomic regions" It also uses local mappability information. I never used it because it has no CLI (only R).