Introduction

Chromatin immunoprecipitation (ChIP) followed by DNA sequencing (ChIP-seq) and ChIP followed by genome tiling array analysis (ChIP-chip) have become prevalent high throughput technologies for identifying the binding sites of DNA-binding proteins genome-wise. A number of algorithms have been published to facilitate the identification of the binding sites of the DNA-binding proteins of interest. The identified binding sites as the list of peaks are usually converted to BED or bigwig files to be loaded to the UCSC genome browser as custom tracks for investigators to view the proximity to various genomic features such as genes, exons or conserved elements. However, clicking through the genome browser is a daunting task when the number of peaks gets large or the peaks spread widely across the genome.

Here we developed ChIPpeakAnno, a Bioconductor1 package, to facilitate the batch annotation of the peaks identified from ChIP-seq or ChIP-chip experiments. We implemented functionality to find the nearest gene, exon, miRNA or other custom features supplied by users such as the most conserved elements and other transcription factor binding sites leveraging GRanges. Since the genome annotation gets updated frequently, we have leveraged the biomaRt package to retrieve the annotation data on the fly. The users also have the flexibility to pass their own annotation data or annotation from GenomicFeatures as GRanges. We have also leveraged BSgenome and biomaRt to retrieve the sequences around the identified peak for peak validation or motif discovery2. To understand whether the identified peaks are enriched around genes with certain GO terms, we have implemented the Gene Ontology (GO) enrichment test in the ChIPpeakAnno package leveraging the hypergeometric test phyper in the stats package and integrated with the GO annotation from the GO.db package and multiplicity adjustment functions from the multtest package3–8. The pathway analysis using reactome or KEGG is also supported. Starting 3.4, we also implement the functions for permutation test to determine whether there is a significant overlap between two sets of peaks. In addition, binding patterns of multiple transcription factors (TFs) or distributions of multiple epigenetic markers around genomic features could be visualized and compared easily using a side-by-side heatmap and density plot.

Quick start

library(ChIPpeakAnno)
## import the MACS output
macs <- system.file("extdata", "MACS_peaks.xls", package="ChIPpeakAnno")
macsOutput <- toGRanges(macs, format="MACS")
## annotate the peaks with precompiled ensembl annotation
data(TSS.human.GRCh38)
macs.anno <- annotatePeakInBatch(macsOutput, AnnotationData=TSS.human.GRCh38)
## add gene symbols
library(org.Hs.eg.db)
macs.anno <- addGeneIDs(annotatedPeak=macs.anno,
                        orgAnn="org.Hs.eg.db",
                        IDs2Add="symbol")

if(interactive()){## annotate the peaks with UCSC annotation
    library(GenomicFeatures)
    library(TxDb.Hsapiens.UCSC.hg38.knownGene)
    ucsc.hg38.knownGene <- genes(TxDb.Hsapiens.UCSC.hg38.knownGene)
    macs.anno <- annotatePeakInBatch(macsOutput,
                                     AnnotationData=ucsc.hg38.knownGene)
    macs.anno <- addGeneIDs(annotatedPeak=macs.anno,
                            orgAnn="org.Hs.eg.db",
                            feature_id_type="entrez_id",
                            IDs2Add="symbol")
}

An example of ChIP-seq analysis workflow using ChIPpeakAnno

We illustrate here a common downstream analysis workflow for ChIP-seq experiments. The input of ChIPpeakAnno is a list of called peaks identified from ChIP-seq experiments. The peaks are represented by GRanges in ChIPpeakAnno. We implemented a conversion functions toGRanges to convert commonly used peak file formats, such as BED, GFF, or other user defined formats such as MACS (a popular peak calling program) output file to GRanges. Please type ?toGRanges for more information.

The workflow here exemplifies converting the BED and GFF files to GRanges, finding the overlapping peaks between the two peak sets, and visualizing the number of common and specific peaks with Venn diagram.

bed <- system.file("extdata", "MACS_output.bed", package="ChIPpeakAnno")
gr1 <- toGRanges(bed, format="BED", header=FALSE)
## one can also try import from rtracklayer
library(rtracklayer)
gr1.import <- import(bed, format="BED")
identical(start(gr1), start(gr1.import))
## [1] TRUE
gr1[1:2]
## GRanges object with 2 ranges and 1 metadata column:
##               seqnames      ranges strand |     score
##                  <Rle>   <IRanges>  <Rle> | <numeric>
##   MACS_peak_1     chr1 28341-29610      * |    160.81
##   MACS_peak_2     chr1 90821-91234      * |    133.12
##   -------
##   seqinfo: 1 sequence from an unspecified genome; no seqlengths
gr1.import[1:2] #note the name slot is different from gr1
## GRanges object with 2 ranges and 2 metadata columns:
##       seqnames      ranges strand |        name     score
##          <Rle>   <IRanges>  <Rle> | <character> <numeric>
##   [1]     chr1 28341-29610      * | MACS_peak_1    160.81
##   [2]     chr1 90821-91234      * | MACS_peak_2    133.12
##   -------
##   seqinfo: 1 sequence from an unspecified genome; no seqlengths
gff <- system.file("extdata", "GFF_peaks.gff", package="ChIPpeakAnno")
gr2 <- toGRanges(gff, format="GFF", header=FALSE, skip=3)
ol <- findOverlapsOfPeaks(gr1, gr2)
makeVennDiagram(ol,
                fill=c("#009E73", "#F0E442"), # circle fill color
                col=c("#D55E00", "#0072B2"), #circle border color
                cat.col=c("#D55E00", "#0072B2"))