RJunBase: a database of RNA splice junctions in human normal and cancerous tissues

Qin Li, Hongyan Lai, Yuchen Li, Bing Chen, Siyuan Chen, Yan Li, Zhaohui Huang, Zhiqiang Meng, Peng Wang, Zhixiang Hu, Shenglin HuangNucleic Acids Res 2021 · DOI

TL;DR

Researchers have developed RJunBase, a comprehensive web-accessible database that integrates and annotates three types of RNA splice junctions (linear, back-splice, and fusion) from both normal and cancerous human tissues. This resource aims to provide a unified platform for exploring splicing patterns, identifying tumor-specific junctions, and facilitating the discovery of novel diagnostic and therapeutic targets in cancer.

What Problem Does This Paper Address?

The intricate process of RNA splicing is fundamental to generating diverse protein isoforms from a single gene, and its dysregulation is increasingly implicated in human cancers. While existing databases offer insights into specific aspects of RNA splicing, there's a notable lack of a consolidated resource that comprehensively catalogs and visualizes various splice junction types (linear, back-splice, and fusion) across a wide range of human normal and cancerous tissues. This gap hinders researchers from easily identifying and analyzing tumor-specific splicing events, which hold significant potential for cancer diagnosis and treatment.

Methods

RJunBase was constructed by integrating data from multiple sources. Linear junctions were identified from RNA-seq data of 10,283 TCGA samples and the GTEx portal using the STAR alignment and StringTie assembly tools, followed by quantification and normalization with the ASJA software package. Fusion junctions were obtained from a previous study utilizing the DEEPEST algorithm on TCGA RNA-seq data. Back-splice junctions were sourced from the MiOncoCirc database. The database employs a MySQL backend and a Twirl template engine for its web interface, with a Scala script for querying. Novel visualization techniques include a circular layout for fusion junctions and boxplots for expression profiles, alongside Kaplan-Meier curves for survival analysis.

Datasets

The RJunBase database integrates data from:

  • The Cancer Genome Atlas (TCGA): Used for identifying linear and fusion junctions from 10,283 samples across 33 cancer types.
  • Genotype-Tissue Expression (GTEx) portal: Provided linear junction data from 17,382 normal tissue samples.
  • MiOncoCirc database: Contributed back-splice junction data from 1,959 samples.

The database also includes annotations from the GENCODE database (release 29) and incorporates information from ONGene and TSGene for tumor-associated gene identification.

Key Insights and Ideas

  • Unified Resource for Splice Junctions: RJunBase addresses the fragmentation of splice junction data by bringing together linear, back-splice, and fusion junctions into a single, searchable platform.
  • Identification of Tumor-Specific Junctions: The database is designed to highlight junctions that are uniquely expressed or altered in cancer tissues, which are prime candidates for diagnostic markers and therapeutic targets.
  • Visualization of Splicing Patterns: Interactive tools allow users to visualize gene-level splicing patterns and junction-level expression profiles, offering a deeper understanding of splicing complexity.
  • Potential for Neoepitope Discovery: The inclusion of splice junctions, particularly those unique to tumors, opens avenues for identifying novel neoepitopes for cancer immunotherapy.
  • Comprehensive Annotation: Junctions are annotated with gene information, tumor association status, and expression data, facilitating downstream analysis.

Key Results

The initial release of RJunBase contains:

  • 682,017 linear junctions
  • 225,949 back-splice junctions
  • 34,733 fusion junctions
  • Data from 18,084 non-cancerous and 11,540 cancerous samples.
  • 324,651 unannotated junctions and 32,837 tumor-specific junctions.

The paper demonstrates the utility of RJunBase through case studies, showing how users can explore the expression and prognostic significance of specific junctions, such as ETV4_LS002 and UN_SLC22A10_LS0017, in various cancer types. For instance, high expression of ETV4_LS002 was correlated with decreased overall survival in adrenocortical carcinoma and liver hepatocellular carcinoma.

Limitations and Future Work

The authors acknowledge that some unannotated junctions might be artifacts of bioinformatics analysis. Future work will focus on continuously updating RJunBase with more splice junction data from healthy and pathological human tissues, especially rare cancers, to expand its utility as a resource for understanding cancer transcriptomes and identifying new diagnostic and therapeutic targets. The database is also designed to allow researchers to submit new splice junction data, fostering community contribution.

Summary updated

BibTeX
@article{li2021rjunbase,
  author = {Qin Li and Hongyan Lai and Yuchen Li and Bing Chen and Siyuan Chen and Yan Li and Zhaohui Huang and Zhiqiang Meng and Peng Wang and Zhixiang Hu and Shenglin Huang},
  title = {{RJunBase: a database of RNA splice junctions in human normal and cancerous tissues}},
  journal = {Nucleic Acids Res.},
  year = {2021},
  doi = {10.1093/nar/gkaa1056},
}
Figure