SEanalysis: a web tool for super-enhancer associated regulatory analysis

Fengcui Qian, Xuecang Li, Jincheng Guo, Jianmei Zhao, Yanyu Li, Zhidong Tang, Liwei Zhou, Jian Zhang, Xuefeng Bai, Yong Jiang, Qi Pan, Qiuyu Wang, Enmin Li, Chunquan Li, Liyan Xu, Dechen LinNucleic Acids Res 2019 · DOI

TL;DR

This paper introduces SEanalysis, a novel web server designed to analyze the complex regulatory networks associated with super-enhancers (SEs). By integrating diverse datasets, SEanalysis provides a comprehensive platform for researchers to explore how SEs, transcription factors (TFs), and pathways interact to regulate gene expression, offering crucial insights into cellular identity and disease mechanisms.

What Problem Does This Paper Address?

Super-enhancers (SEs) are critical regulatory elements that drive cell-type-specific gene expression programs, playing significant roles in both normal biological processes and diseases like cancer. While several databases exist to catalog SEs, they largely lack the functionality to analyze the downstream and upstream regulatory relationships involving these elements. Understanding these complex networks, which involve interactions between SEs, the transcription factors (TFs) that bind to them, and the signaling pathways they influence, is crucial for deciphering gene regulation. This paper addresses this gap by developing SEanalysis, a web tool that provides comprehensive network analyses of SEs, TFs, pathways, and SE-associated genes.

Methods

SEanalysis integrates several key components and analytical approaches:

  • SE Annotation: SEs were identified using H3K27ac ChIP-seq data and the ROSE software, a standard method for SE identification. Four strategies were employed to annotate SE-associated genes: closest active genes, overlapping genes, proximal genes, and closest genes.
  • TF Occupancy Identification: A large collection of TF ChIP-seq datasets (over 5000) from various sources were processed and mapped to the hg19 genome. TF binding peaks overlapping with SEs were identified using BEDTools. Additionally, TF binding motifs for approximately 700 human TFs were used with FIMO to predict TF occupancy in SE regions, allowing for analysis even when direct ChIP-seq data might be limited.
  • Master TF Identification and TF Classification: The CRC_Mapper program was used to identify master TFs (auto-regulated TFs encoded by SE-associated genes that form auto-regulatory loops). TFs were also classified into superclass, class, family, and subfamily categories based on their DNA-binding domains, using the TFClass database.
  • Network Construction: Relationships were established between SEs and TFs (via ChIP-seq or motif analysis), between TFs and pathways (if the TF is a pathway component), and between SEs and genes. These relationships were merged to construct comprehensive SE-associated regulatory networks.
  • Core Analysis Functions: SEanalysis offers three main analytical modules:
  • Pathway Downstream Analysis: Identifies enriched pathways for a given set of genes, then determines downstream TFs, SEs occupied by these TFs, and SE-associated genes. This involves hypergeometric tests for pathway enrichment and FDR correction for multiple testing.
  • Upstream Regulatory Analysis: Identifies SEs associated with input genes, the TFs occupying these SEs, and the upstream pathways of these TFs.
  • Genomic Region Annotation: Allows users to input genomic regions (in BED format) to identify overlapping SEs, associated genes, TFs, and upstream pathways.
  • Interactive Visualization: The tool provides interactive visualization of regulatory networks using Echarts and a customizable genome browser (JBrowse) with over 6,000 tracks for visualizing SEs, TFBS, SNPs, DHSs, and conservation scores.

Datasets

SEanalysis integrates data from numerous public resources:

  • Super-enhancers (SEs): Over 330,000 SE regions from more than 540 cell/tissue types, sourced from SEdb (which itself uses data from NCBI GEO/SRA, ENCODE, and Roadmap Epigenomics).
  • TF ChIP-seq Data: 5042 TF ChIP-seq datasets from ENCODE, Remap, Cistrome, ChIP-Atlas, and GTRD.
  • TF Binding Motifs: DNA-binding sequence motifs for approximately 700 human TFs, compiled from TRANSFAC, MEME suite, JASPAR CORE, Jolma2013, Homeodomains, UniPROBE, and Wei2010.
  • Pathway Databases: 2880 pathways from 10 databases: KEGG, Reactome, NetPath, WikiPathways, PANTHER, PID, HumanCyc, CTD, SMPDB, and INOH.
  • Genomic Information: Reference genomes (hg19), and links to external resources like NCBI Gene, GeneCards, and UniProt.

The paper does not introduce new benchmark datasets but rather integrates existing ones for comprehensive analysis.

Key Insights and Ideas

  • Bridging the Gap in SE Analysis: The primary contribution is moving beyond simple SE cataloging to providing a functional, network-based analysis of SEs. This allows researchers to understand how SEs contribute to regulatory processes.
  • Integrated Network View: SEanalysis offers a holistic view by connecting pathways, TFs, SEs, and genes, revealing complex regulatory cascades that are difficult to discern from individual data types.
  • Predictive Power for TF-SE Interactions: The combination of ChIP-seq data and motif analysis allows for robust identification of TF binding to SEs, even in cases where direct experimental data might be sparse.
  • Uncovering Master Regulators: The identification of master TFs within SE-associated regulatory circuits provides insights into key drivers of cellular identity and potential therapeutic targets.
  • Customizable and Interactive Exploration: The web server's user-friendly interface, interactive visualizations, and customizable genome browser empower researchers to explore complex data in an intuitive manner.

Key Results

The paper demonstrates the utility of SEanalysis through case studies:

  • Pathway Downstream Analysis Validation: In a colon cancer cell line (HCT116) treated with Wnt modulators, SEanalysis successfully identified the Wnt signaling pathway as significantly enriched among differentially expressed genes. It also pinpointed key downstream TFs like TCF7L2, which were shown to occupy a vast majority of HCT116 SEs, aligning with previous experimental findings.
  • Upstream Regulatory Analysis Validation: For luminal breast cancer, SEanalysis predicted SEs associated with the ESR1 gene being occupied by estrogen receptor alpha (ERa). It also identified relevant pathways like "Nuclear receptor transcription pathway," reinforcing the known role of ERa in mediating estrogen signaling and regulating SE activity.

These case studies validate the core functions of SEanalysis, demonstrating its ability to accurately predict pathway enrichment, TF-SE associations, and upstream regulatory relationships.

Limitations and Future Work

  • Data Coverage and Updates: While extensive, the SE and TF datasets are snapshots in time. Future work will involve updating these datasets as new data becomes available to maintain the server's relevance.
  • Computational Complexity: Analyzing large-scale regulatory networks can be computationally intensive. The authors have implemented efficient algorithms, but further optimization might be needed for extremely large user inputs.
  • Interpretation of "Master TFs": The definition and identification of master TFs are based on specific criteria (auto-regulation, interconnected loops). While powerful, alternative definitions or broader interpretations of core regulatory roles could be explored.
  • Future Directions: The authors plan to continuously update and maintain SEanalysis, potentially expanding its analytical capabilities and integrating more data types to further enhance its utility for epigenomic research. The focus on integrative analysis of epigenomic regulatory networks is highlighted as a critical area for future development in the field.

Summary updated

BibTeX
@article{qian2019seanalysis,
  author = {Fengcui Qian and Xuecang Li and Jincheng Guo and Jianmei Zhao and Yanyu Li and Zhidong Tang and Liwei Zhou and Jian Zhang and Xuefeng Bai and Yong Jiang and Qi Pan and Qiuyu Wang and Enmin Li and Chunquan Li and Liyan Xu and Dechen Lin},
  title = {{SEanalysis: a web tool for super-enhancer associated regulatory analysis}},
  journal = {Nucleic Acids Res.},
  year = {2019},
  doi = {10.1093/nar/gkz302},
}
Figure