ReMap 2018: an updated atlas of regulatory regions from an integrative analysis of DNA-binding ChIP-seq experiments

Abstract

With this latest release of ReMap (http://remap.cisreg.eu), we present a unique collection of regulatory regions in human, as a result of a large-scale integrative analysis of ChIP-seq experiments for hundreds of transcriptional regulators (TRs) such as transcription factors, transcriptional co-activators and chromatin regulators. In 2015, we introduced the ReMap database to capture the genome regulatory space by integrating public ChIP-seq datasets, covering 237 TRs across 13 million (M) peaks. In this release, we have extended this catalog to constitute a unique collection of regulatory regions. Specifically, we have collected, analyzed and retained after quality control a total of 2829 ChIP-seq datasets available from public sources, covering a total of 485 TRs with a catalog of 80M peaks. Additionally, the updated database includes new search features for TR names as well as aliases, including cell line names and the ability to navigate the data directly within genome browsers via public track hubs. Finally, full access to this catalog is available online together with a TR binding enrichment analysis tool. ReMap 2018 provides a significant update of the ReMap database, providing an in depth view of the complexity of the regulatory landscape in human.

Figure 1.

Overview of the ReMap database expansion. (A) Analyzed datasets growth in ReMap 2018 compared to ReMap 2015. (B) Evolution of the number of datasets per TRs, ranked across common between both ReMap versions. (C) Common TRs between Public and ENCODE sets of data (gray). Direct comparison of Public and ENCODE repertoire, defined as percentages (%), and as number (Nb) of peaks. (D) Genome coverage fraction of each ReMap dataset (NR non-redundant, CRM Cis Regulatory Modules). (E) Comparison of DNase I-accessible regulatory regions against the ReMap 2018, regions from the Roadmap Epigenomics Consortium defining promoter-only, enhancer-only or enhancer–promoter alternating states (Dyadic). Each dot represents the fraction overlap with ReMap 2018 for one of the 111 epigenomes. (F) Comparison of the Roadmap Epigenomics Consortium chromatin states annotations against the ReMap 2018 catalog, using the Core 15 chromatin states model, and a minimum overlap of 50% between regions. Each dot represents the overlap for one of the 111 epigenomes. Chromatin state definitions and abbreviations are as follows; 1 Active TSS (TssA), 2 Flanking active TSS (TssAFlnk), 3 Transcr. at gene 5′ and 3′(TxFlnk), 4 Strong transcription (Tx), 5 Weak transcription (TxWk), 6 Genic enhancers (EnhG), 7 Enhancers (Enh), 8 ZNF genes + repeats (ZNF/Rpts), 9 Heterochromatin (Het), 10 Bivalent/poised TSS (TssBiv), 11 Flanking bivalent TSS/Enh (BivFlnk), 12 Bivalent enhancer (EnhBiv), 13 Repressed Polycomb (ReprPC), 14 Weak repressed Polycomb (ReprPCWk), 15 Quiescent/low (Quies).

Figure 2.

ReMap ChIP-seq binding pattern of 2829 datasets. A genome browser example of the ChIP-seq binding peak depth of the ReMap 2018 catalog compared to ReMap 2015 at the vicinity of the ELAC1 promoter (chr18:50,967,094-50,970,983). The tracks and peaks displayed are compacted to thin lines so the depth of ReMap 2018 bindings can be compared to ReMap 2015. A full and un-compacted screenshot is available as Supplementary Figures S2 and 3. On this location the ReMap 2018 catalog contains 1307 peaks, whereas the ReMap 2015 contains 229 peaks (ReMap 2015 lifted to GRCh38/hg38 assembly). The following genome tracks correspond to the GENCODE v24 Comprehensive Transcript Set and the 100 vertebrates base-wise conservation showing sites predicted to be conserved (positive scores in blue), and sites predicted to be fast-evolving (negative scores in red). A detailed view of the redundant peaks for a FOXA1 site is available in Figure 3.

Figure 3.

FOXA1 ChIP-seq peaks pattern evolution across ReMap versions. Detailed view of the FOXA1 peaks present in ReMap 2018 (60 peaks) compared to the FOXA1 peaks in ReMap 2015 (15 peaks) found at the genomic location chr18:50,969,638-50,970,931 in the first intron of the ELAC1 gene. Those 60 FOXA1 peaks are derived from GEO, ArrayExpress and ENCODE ChIP-seq across multiple cell lines. Interestingly, it can be noted that the peak summits (vertical bars) of each peak aggregate closely from each other, defining precisely the DNA binding location. Those aggregations of the FOXA1 summits are an illustration of what is globally observed for peaks of different TFs across the genome.