ChIP-Atlas 2021 update: a data-mining suite for exploring epigenomic landscapes by fully integrating ChIP-seq, ATAC-seq and Bisulfite-seq data

1In the past decade, despite the increasing number of high-throughput sequencing experiments, the secondary use of the obtained raw data has required complex and large-scale computational processing, and thus most data are still being hoarded. Since 2015, we have been comprehensively collecting, analyzing and integrating almost all chromatin immunoprecipitation sequencing (ChIP-seq) and DNase-seq—a method for profiling chromatin regions accessible to DNase—data derived from six representative model organisms (Homo sapiens, Mus musculus, Rattus norvegicus, Drosophila melanogaster, Caenorhabditis elegans and Saccharomyces cerevisiae) archived at the Sequence Read Archive (SRA). SRA is the largest publicly available data repository that accepts submissions of high-throughput sequencing data, which is maintained by NCBI, EBI, and DDBJ. Data-mining tools powered by these data were provided through our web server, ChIP-Atlas (https://chip-atlas.org), for visualization of assembled peak data and enrichment analysis for given genomic loci to identify transcription factor (TF) binding and histone modification status. ChIP-Atlas is powered by dedicated manual curation and annotation of experimental metadata, and a uniform data process pipeline to reveal the complex architecture of the transcription regulatory landscape.

2Information on TF binding and histone marks is, however, still insufficient to fully understand the regulatory systems for gene expression control because an absence of nucleosomes and low methylation levels also characterize active promoters, enhancers, and other gene regulatory sequences. To detect accessible chromatin regions, several experimental methods including DNase-seq, formaldehyde-assisted isolation of regulatory elements, and assay for transposase-accessible chromatin with sequencing (ATAC-seq) have been developed. Among these, ATAC-seq identifies accessible chromatin regions based on their increased accessibility to Tn5 transposase integration and is now predominantly used given its advantages of technical ease and sensitivity to a small number of cells. In contrast, bisulfite sequencing (Bisulfite-seq) is a well-established protocol to detect methylated cytosines in genomic DNA, which employs a chemical method that selectively deaminates unmodified cytosine to uracil while leaving 5-methylcytosine intact before DNA sequencing.

3Here, we describe the ChIP-Atlas 2021 update, which collected all the ATAC-seq (n = 66 104) and whole-genome bisulfite sequencing (WGBS) data (n = 51 074) for the six representative model organisms. This update makes ChIP-Atlas, to our knowledge, the first and only web service that enables users to explore not only protein binding, but also chromatin accessibility and DNA methylation status, within a single given genomic region of interest (ROI) at the same time (Peak Browser tool). ChIP-Atlas can also be used to reveal the regulatory network involved in a batch of genomic ROIs (Enrichment Analysis tool) based on ATAC-seq and WGBS data in addition to ChIP-seq data in the previous version. ChIP-Atlas provides a panoramic view of the whole epigenomic landscape and should be of great interest to researchers in the fields of genetics and genomics, as well as those studying transcriptional regulation in general.

4We obtained sample metadata from the NCBI BioSample database FTP server (ftp://ftp.ncbi.nlm.nih.gov/biosample). We also retrieved the metadata of SRA experiments (accession number with the prefix SRX, ERX or DRX; hereafter SRXs), such as library preparation or used sequencing instrument models, from the NCBI SRA FTP server (ftp://ftp.ncbi.nlm.nih.gov/sra/reports/Metadata). ChIP-Atlas uses ATAC-seq and WGBS SRXs that meet the following criteria: LIBRARY_STRATEGY is ‘ATAC-seq’ or ‘Bisulfite-seq’; LIBRARY_SOURCE is ‘GENOMIC’; LIBRARY_SELECTION is ‘PCR’ or ‘others’ for ATAC-seq, and ‘RANDOM’ for WGBS; taxonomy_name is ‘Homo sapiens,’ ‘Mus musculus’, ‘Rattus norvegicus’, ‘Caenorhabditis elegans’, ‘Drosophila melanogaster’ or ‘Saccharomyces cerevisiae’; and INSTRUMENT_MODEL includes ‘Illumina’, ‘NextSeq’ or ‘HiSeq.’ For ChIP-seq and DNase-seq data, refer to the previous paper on ChIP-Atlas.

5Binarized sequence raw data (.sra) for each SRX were downloaded and decompressed into FASTQ format with the ‘fasterq-dump’ command of SRA Toolkit (ver. 2.9.4; https://ftp-trace.ncbi.nlm.nih.gov/sra/sdk/2.9.4/sratoolkit.2.9.4-ubuntu64.tar.gz) according to the default mode, with the exception of paired-end reads, which were decoded with the ‘-split-files’ option. For ATAC-seq data, the subsequent alignment (with Bowtie2 [ver. 2.2.2]) and peak call (with MACS2 [ver. 2.1.1; macs2 callpeak]) process was the same as that for ChIP-seq and DNase-seq data, which was previously described in the first ChIP-Atlas paper. As for WGBS data, BMap (ver. 1.0; https://github.com/FumihitoMiura/Project-2/blob/master/Project-2.tar.gz), a speedy aligner with small temporary file sizes, was used for alignment and MethPipe (ver. 4.1.1; https://github.com/smithlabcode/methpipe/releases/download/v4.1.1/methpipe-4.1.1.tar.gz) was subsequently used as a region-caller for identifying hyper-, partially and hypo-methylated regions (MRs). After creation of the index files for all genome assemblies, we aligned the downloaded FASTQ files against the reference genomes with the ‘-fastq’ and ‘-pfastq’ options for single-end and paired-end reads, respectively. The alignment data (in bigWig format) containing methylation level and coverage information were generated for each SRX. Methylation level and coverage were then calculated for each CpG, and these data were next streamed into MethPipe. The ‘hypermr,’ ‘pmd,’ and ‘hmr’ sub-commands of MethPipe were used for identifying hyper-, partially, and hypo-MRs (in bigBed format), respectively, for each SRX according to the default mode.

6GWAS SNP data were downloaded from the website of UCSC genome browser (https://hgdownload.soe.ucsc.edu/goldenPath/hg19/database/gwasCatalog.txt.gz; on 21 December 2021). A BED format file was created by extending the genomic coordinates of all SNPs by 5 kb up- and downstream. Genomic regions containing SNPs were then categorized into ‘inflammatory bowel disease (IBD)-associated’ (n = 393) and ‘Others’ (n = 189 036) according to the disease/phenotype names listed in the ‘trait’ column. The IBD-associated SNPs (in BED format) were directly used as ‘dataset A’ for the Enrichment Analysis tool on the ChIP-Atlas web interface. Meanwhile, 3930 SNPs in the ‘Others’ category, which is ten times the number of IBD-associated SNPs, were randomly selected as the background data for ‘dataset B’.

7In this update, the major expansion of the experiment types results in significant growth in the number of experiments and the number of annotated functional genome regions by including all ATAC-seq and WGBS datasets archived at SRA (Figure 1A, Table 1). For unified management of the records, each experiment is assigned a unique ID (hereafter referred to as ‘SRX’) in ChIP-Atlas, which is exactly the same as the original SRA accession number. The number of SRXs collected in ChIP-Atlas is over 300 000 for the six organisms, which corresponds to 84.3% of the total number of ChIP-seq, ATAC-seq, DNase-seq and Bisulfite-seq SRXs in SRA for all of these organisms (n = 362 121 as of September 2021; Table 1). In this update, we adopted the latest version of genome assemblies in addition to previous ones (Table 2). Since the public release of ChIP-Atlas, the data have been updated monthly concurrent with the monthly update of the NCBI SRA metadata dump (Figure 1B), by which >100 000 ChIP-seq experimental datasets have been added since 2018 (Table 1).

8We manually curated and annotated the cell types used in each experiment according to commonly or officially adopted nomenclature. The cell types were further categorized into superordinate ‘cell type classes’ (Figure 1C). We adapted a uniformed data process pipeline (Figure 1D; detailed in the Materials and Methods), in which the ChIP-seq, ATAC-seq, and DNase-seq data are aligned to corresponding reference genomes with Bowtie2 and then subjected to peak calling with MACS2; meanwhile, for WGBS data, we align the raw reads against the reference genomes using BMap before applying MethPipe for statistically calling hyper-, partially, and hypo-methylated regions (MRs). All alignment and peak-call data are freely downloadable from http://dbarchive.biosciencedbc.jp/kyushu-u/ (detailed on the documentation page of ChIP-Atlas [https://github.com/inutano/chip-atlas/wiki/#downloads_doc]), and can be browsed by users in the IGV genome browser by entering the SRX ID or a given keyword (or keywords) in the corresponding Data Search page of ChIP-Atlas (Supplementary Figure S1 and Supplementary Material S1).

9All peak-call data recorded in ChIP-Atlas can be presented graphically with the Peak Browser tool. One can therefore easily understand not only protein–genome interactions (ChIP-seq), but also chromatin accessibility (ATAC-seq) and DNA methylation levels (WGBS), within any query genomic region of interest (ROI). To implement this tool, we integrated a large amount of peak-call data, indexed them for IGV, and constructed a web interface that externally controls IGV preinstalled on the user's machine (tested on Mac, Windows and Linux platforms). For instance, upon the specification of ChIP-seq, ATAC-seq or WGBS data of mouse spermatogonia on the web page (Figure 2A), the corresponding results are automatically streamed into IGV (Figure 2B). In this case, multiple ATAC-seq and WGBS in spermatogonia data characterize an accessible chromatin region and hypo-MR, respectively, at the locus between the Tcf19 and Cchcr1 genes, where multiple factors such as Smarca4, Zbtb16, and Sall4 are colocalized, suggesting that the Tcf19–Cchcr1 locus is robustly hypo-methylated and open to bind Sall4 and other TFs in spermatogonia. Representative ChIP-seq (SRX1284250), ATAC-seq (SRX5884282) and WGBS (SRX749893) alignment data are also exhibited in the top panel of Figure 2B. The IGV sessions can be saved as XML files at any moment, so that the results obtained by using the Peak Browser tool can be easily and precisely shared among collaborators (Supplementary Material S2). With the use of ChIP-Atlas, users can not only check individual experimental data, but also browse an integrative landscape of multiple epigenetic profiling results, potentially providing useful insight into the location of functional genomic regions (enhancers, promoters and insulators) and the corresponding regulators.

10Enrichment Analysis is a tool that allows a search for TFs, accessible chromatin regions, and hypo- or hyper-MRs enriched at a batch of genomic ROIs or gene loci. Upon the submission of two sets of genomic regions (ROIs and background regions), all SRXs are evaluated to count the overlaps between the peaks and submitted regions and to perform Fisher's exact test, before returning the enrichment analysis results in HTML and TSV formats. Although ChIP-Atlas can generate random background regions for comparison to ROIs, the users are strongly recommended to provide biologically appropriate background genomic intervals because most randomly selected regions are probabilistically devoid of any functional genomic annotation. The results are assigned unique URLs, which are permanently available to the public. As an example of usage, we selected inflammatory bowel disease (IBD)-associated SNPs identified by GWAS as the ROIs (n = 393) and other SNPs as the background (n = 3930), and we applied these selections to the Enrichment Analysis tool (Figure 2C; detailed in the Materials and Methods). The results in HTML format are shown in Figure 2D (upper), including SRX IDs (column 1), features (column 3), cell types (column 5), P-values (column 9) and fold enrichments (column 11), where hypo-MRs of neutrophils are significantly enriched (P = 1 × 10−17). To visually interpret the results, we downloaded the resultant TSV files (refer to Supplementary Table S1 for unique URLs for each analysis) and generated volcano plots (Figure 2D [bottom]; Supplementary Table S1). The IBD-associated SNPs were shown to be enriched in hypo-MRs and accessible chromatin regions of blood-related cells (red in Figure 2D), including the ATAC-seq peaks of macrophages (P = 1 × 10−20) and Th17 cells (P = 1 × 10−17), both involved in the inflammation of IBD. Furthermore, the enrichment analysis of TF ChIP-seq data revealed that the IBD-associated SNPs were preferentially bound by STAT1 in monocytes (P = 1 × 10−22) and SPI1 in macrophages (P = 1 × 10−21), essential factors for inflammation and macrophage differentiation, respectively, which is consistent with the nature of IBD as an autoimmune disease. All resultant URLs for generating Figure 2D are summarized in Supplementary Table S2. API of the Enrichment Analysis tool is also provided by ChIP-Atlas, the general instructions for which can be found on the documentation page (https://github.com/inutano/chip-atlas/wiki/Perform-Enrichment-Analysis-programmatically).

11In this paper, we present a major update of ChIP-Atlas involving significant expansion in the number of experiments by including all public ATAC-seq and WGBS data. The updated web service enables users to explore not only protein binding, but also chromatin accessibility and DNA methylation levels within single (Peak Browser tool) or multiple (Enrichment Analysis tool) queries of genomic ROIs or gene loci. As an example of use, we performed enrichment analysis to reveal that IBD-associated SNPs are the most significantly hypo-methylated in the neutrophils, a granulocyte subtype known to be involved in autoinflammatory IBD.

12Before and after the public release of ChIP-Atlas, several similar web services have been released (Table 2). For ChIP-seq, ATAC-seq and DNase-seq data, Cistrome DB (http://cistrome.org/db/), ReMap (https://remap2022.univ-amu.fr/), and GTRD (https://gtrd.biouml.org/) are representatives providing thousands of preprocessed datasets. As for Bisulfite-seq data, MethBank (http://bigd.big.ac.cn/methbank) is widely used. ChIP-Atlas covers much more experimental data than all of these other services. ChIP-Atlas does not cover ChIP-exo and MNase-seq data, in contrast to ReMap and GTRD, and there are fewer available organisms than in GTRD and MethBank. Alignment data (in bigWig format) are available from ChIP-Atlas, Cistrome DB and MethBank. Integrative analysis tools are provided in all services. Enrichment analysis is possible with ChIP-Atlas in both a GUI and an API, while ReMap provides such a tool in a CLI (R package named ‘ReMapEnrich’). Since the publication of the first ChIP-Atlas paper, we have improved the service to make it compatible with the latest versions of reference genomes such as hg38 for human and mm10 for mouse, as is the case for other services. Meanwhile, alignment of raw reads against old references is still performed during the monthly update of ChIP-Atlas, and alignment and peak-call data in old versions are still provided for analyzing the data of users obtained years ago.

13In addition to the examples of use mentioned in the RESULTS section, ChIP-Atlas has been cited by hundreds of peer-reviewed articles since its first release in 2015, including research for analyzing cis-regulatory elements of certain genes, and TF enrichment at genomic ROIs and query genes (see http://chip-atlas.org/publications for the full list of publications citing ChIP-Atlas). Furthermore, because all alignment (bigWig) and peak-call (bigBed) data can be freely downloaded, ChIP-Atlas is now interconnecting with many other databases or web services such as UCSC Browser, DeepBlue (an epigenomic data server providing a central data access hub for large collections of epigenomic data), RegulatorTrail (a web service predicting target genes of TFs), jPOSTrepo (a data repository of sharing raw/processed mass spectrometry data), and the Signaling Pathways Project (a multi-omics knowledgemine based upon public transcriptomic and cistromic datasets). Along with the inclusion of ATAC-seq and WGBS data, and the ongoing monthly updates with semiautomatic pipelines and systematic curation, the source data in ChIP-Atlas are continuously expanding. We are planning to include more experiment types such as CUT&Tag and ChIL-seq and more organisms including plants such as Arabidopsis thaliana. Integration of preprocessed 3D genome conformation data such as Hi-C datasets into the Peak Browser and Enrichment Analysis tool is also on the agenda.

14ChIP-Atlas (https://chip-atlas.org) is a publicly available web server with no sign up required. Documentation for data processing and downloadable data are available in the ‘Documentation’ section (https://github.com/inutano/chip-atlas/wiki).