TL;DR
TransmiR v2.0 is a significantly expanded database that now offers over 3,700 manually curated transcription factor (TF)-microRNA (miRNA) regulations and over 1.7 million TF-miRNA regulations derived from ChIP-seq data across 19 species. This updated resource aims to provide a more comprehensive and user-friendly platform for investigating the complex regulatory networks involving TFs and miRNAs, crucial for understanding gene expression and disease mechanisms.
What Problem Does This Paper Address?
Understanding how transcription factors (TFs) regulate microRNAs (miRNAs) is fundamental to deciphering gene expression control, especially in the context of various biological processes and diseases. While miRNAs are known to be critical post-transcriptional regulators, the intricate mechanisms by which TFs control their expression are not fully understood. Existing resources for TF-miRNA regulations were limited, either due to difficulties in precisely identifying miRNA promoter regions or a lack of comprehensive literature curation. This paper addresses this gap by presenting TransmiR v2.0, an updated and expanded database that integrates both manually curated literature data and high-throughput experimental data (ChIP-seq) to provide a more complete picture of TF-miRNA regulatory interactions.
Methods
The construction of TransmiR v2.0 involved several key methodological approaches:
- Manual Literature Curation: The team surveyed over 8,000 publications from 2013-2017 to curate TF-miRNA regulations supported by promoter-related experiments (e.g., ChIP, luciferase reporter assays). This process involved standardizing TF and miRNA names, incorporating external annotations (gene IDs, associated diseases, cancer prognostic data), and linking to relevant databases like miRBase and DisGeNET.
- ChIP-seq Data Integration: TF-miRNA regulations were derived from ChIP-seq data for five species (human, mouse, rat, fruit fly, and worm). This involved:
- miRNA Promoter Identification: Defining putative miRNA transcription start sites (TSSs) based on pre-miRNA sequences or host gene 5'-ends, and establishing promoter regions (5 kb upstream to 1 kb downstream). For higher confidence, a refined promoter region (300 bp upstream to 100 bp downstream) was used for TSSs supported by high-throughput experiments.
- ChIP-seq Peak Analysis: Tissue-specific ChIP-seq data from the ChIP-Atlas database was used. Overlaps between TF binding peak regions and miRNA promoter regions were identified to infer TF-miRNA regulations. These were classified into "level 1" (predicted) and "level 2" (supported by high-throughput experimental data).
- TF Binding Motif Prediction: For human TF-miRNA regulations, predicted interactions were generated by scanning miRNA promoters for TF binding motifs obtained from the UCSC genome browser.
- Disease-Specific Network Construction: TF-disease and miRNA-disease associations were integrated to construct networks representing TF-miRNA regulatory relationships within specific disease contexts.
- Enrichment Analysis: A hypergeometric test was employed to identify overrepresented miRNA sets (i.e., regulating TFs) within a user-provided list of miRNAs, allowing for the inference of potential regulators.
- Web Server Functionalities: New modules were developed for "Predict" (querying predicted TF-miRNA regulations), "Network" (visualizing TF-miRNA regulations), and "Enrichment analysis."
Datasets
- Literature-Curated Data: Manually curated TF-miRNA regulations from >8,000 publications.
- ChIP-seq Data: Publicly available ChIP-seq data from the ChIP-Atlas database for five species.
- miRNA Information: miRNA genome coordinate information from miRBase v22.
- TF Binding Motifs: Data from the UCSC genome browser.
- External Databases: DisGeNET (disease-gene associations), Human Protein Atlas (cancer prognostic data), OMA (human orthologs), HMDD v2.0 (miRNA-disease associations), TarBase and miRTarBase (miRNA-target interactions), UCSC RefGene dataset (miRNA host genes), and NCBI (TF official symbols).
Key Insights and Ideas
- Bridging the Gap: TransmiR v2.0 effectively bridges the gap between manually curated literature data and large-scale experimental data (ChIP-seq), offering a more comprehensive and reliable resource for TF-miRNA regulation.
- Hierarchical Evidence Levels: The classification of ChIP-seq derived regulations into "level 1" (predicted) and "level 2" (experimentally supported) provides users with a clear understanding of the evidence supporting each interaction.
- Enhanced Usability: The introduction of interactive "Network" visualization and "Enrichment analysis" modules significantly enhances the utility of the database for hypothesis generation and interpretation of experimental results.
- Disease-Centric View: The ability to construct disease-specific TF-miRNA regulatory networks offers a powerful tool for understanding the molecular underpinnings of various diseases.
Key Results
- TransmiR v2.0 contains 3,730 literature-curated TF-miRNA regulations across 19 species, involving 623 TFs and 785 miRNAs, sourced from 1,349 publications.
- The database also incorporates over 1.7 million TF-miRNA regulations derived from ChIP-seq evidence for five key species.
- The literature-curated data shows a 15-fold increase compared to the previous version (TransmiR v1.0).
- The database provides 1,651,502 level 1 and 134,496 level 2 ChIP-seq derived regulations.
- The "Network" module allows for graphical visualization of TF-miRNA regulations for specific TFs, miRNAs, or diseases.
- The "Enrichment analysis" module can identify potential regulating TFs for a given list of miRNAs, with results adjusted for multiple testing (Bonferroni and FDR).
Limitations and Future Work
- miRNA TSS Accuracy: The definition of miRNA gene TSSs can still suffer from inaccuracies, particularly with high-throughput data. The availability of experimentally supported and tissue-specific miRNA TSS data remains limited.
- Action Type Specification: ChIP-seq derived and predicted TF-miRNA regulations do not specify the action type (activation or repression). Future work may involve integrating transcriptome data to infer expression correlations.
- Benchmarking Literature Data: A lack of a standardized benchmarking method for evaluating the confidence of literature-curated regulations is acknowledged. Future efforts will focus on including more detailed information to improve the reliability of this data.
- Continued Improvement: The authors plan to continue improving and expanding the database by integrating more high-quality datasets and developing more powerful tools.
Summary updated