TL;DR
This paper investigates how transcription factors (TFs) and microRNAs (miRNAs) work together to regulate gene expression. They found that TFs and miRNAs often co-regulate genes in pairs, with TF-TF and miRNA-miRNA pairs being more common than TF-miRNA pairs. They also identified a specific regulatory motif called a "feed-forward loop" that is more prevalent in strongly interacting TF-miRNA pairs, suggesting a mechanism for efficient gene regulation and resource management.
What Problem Does This Paper Address?
Understanding how genes are regulated is fundamental to biology. While we know that transcription factors (TFs) control gene expression by binding to DNA and microRNAs (miRNAs) regulate gene expression by affecting mRNA stability and translation, the complex interplay between these two regulatory systems is not fully understood. This paper aims to explore the extent and nature of combinatorial regulation, where multiple regulators (TFs and miRNAs) work together to control gene expression, and how this coordination impacts cellular processes. Specifically, it investigates whether TF-miRNA interactions are as common as TF-TF or miRNA-miRNA interactions and what specific regulatory motifs emerge from these interactions.
Methods
The researchers constructed a comprehensive regulatory network by combining predicted TF targets from the TRANSFAC database with predicted miRNA targets from the PicTar database. To assess the strength and significance of regulatory associations between pairs of TFs and miRNAs, they employed two main statistical approaches:
- Fisher's Exact Test: This standard statistical test was used to evaluate the significance of the overlap between the sets of predicted targets for each pair of regulators. This method is sensitive to sample size, meaning pairs with many targets are more likely to show significant overlap.
- Bayesian Association Measure: To address the sample-size bias of Fisher's Exact Test and to incorporate the "strength" of an interaction (not just its statistical significance), a novel Bayesian approach was developed. This method estimates the posterior probability that the log Odds Ratio of association between a TF and miRNA exceeds a certain threshold, providing a more balanced assessment of biological coordination.
They also used Gene Ontology (GO) annotations to analyze the functional enrichment of gene sets regulated by these pairs, looking for coordinated regulation of functionally related genes.
Datasets
- TRANSFAC database: Used to identify predicted targets of human transcription factors.
- PicTar miRNA database: Used to identify predicted targets of human microRNAs.
- Refseq database: Used to map gene identifiers and their corresponding promoters.
- NCBI Gene database: Used to link Refseq IDs to gene symbols.
- Gene Ontology (GO) catalogue: Used for functional enrichment analysis of gene sets.
The paper does not introduce a new benchmark dataset but rather integrates existing databases to build a comprehensive regulatory network.
Key Insights and Ideas
- TF-TF and miRNA-miRNA pairs are more common regulators than TF-miRNA pairs: This suggests that simpler, intra-layer regulation might be more prevalent, potentially due to evolutionary mechanisms like gene duplication of shorter DNA sequences.
- Master TF regulators exist: A subset of TFs with a very large number of targets appear to co-regulate genes in coordination with a wide range of miRNAs, acting as central hubs in regulatory networks.
- Bayesian approach reveals different interaction patterns: The Bayesian method, which considers both significance and effect size, identified TF-miRNA interactions that were not necessarily driven by regulators with a large number of targets, highlighting potentially important but less obvious regulatory relationships.
- Feed-forward loops (FFLs) are enriched in strongly interacting TF-miRNA pairs: This specific regulatory motif, where a TF activates a gene and a miRNA simultaneously suppresses both the TF and the gene, appears to be a significant mechanism for efficient gene regulation.
- FFLs may prevent cellular resource waste and accelerate mRNA degradation: The proposed function of these FFLs is to ensure rapid and precise control of gene expression, avoiding contradictory processes and dampening fluctuations.
- miRNAs might play roles in cell-cell adhesion and cancer metastasis: The enrichment of certain miRNAs in cell adhesion-related GO terms suggests potential roles in these processes, with some of these miRNAs being implicated in cancer.
Key Results
- Approximately 5.5% of TF-miRNA pairs showed significant co-regulation based on Fisher's Exact Test (q-value < 0.01).
- The Bayesian approach revealed a more uniform distribution of significant TF-miRNA associations, less biased by the number of targets.
- Gene sets associated with cell adhesion were enriched by a significant number of miRNAs, with some of these miRNAs also found to be dysregulated in cancers.
- The proportion of feed-forward loops (FFLs) among strongly interacting TF-miRNA pairs (Fisher's Exact Test q-value < 0.01) was significantly higher (p~1.86 × 10⁻⁹) compared to less interacting pairs.
Limitations and Future Work
- Reliance on predicted targets: The study's conclusions are based on predicted targets of TFs and miRNAs, which have inherent inaccuracies (estimated false positive rates of 50-80%). The authors suggest that analyzing functionally related gene sets can help mitigate this.
- Parameter sensitivity: The construction of the TF binding network involved choosing specific parameters, and the authors acknowledge that varying these parameters could influence the results. They performed robustness checks by exploring different parameter ranges.
- Focus on pairs: The study primarily focuses on pairwise interactions. Future work could explore higher-order combinatorial regulation involving more than two regulators.
- Experimental validation: While the study provides statistical evidence, experimental validation of the identified TF-miRNA interactions and FFLs is crucial.
- Broader functional roles of miRNAs: The paper hints at unknown roles for miRNAs, such as in cell-cell adhesion, suggesting further investigation into these areas.
- Denoising mechanisms: The authors propose that FFLs involving miRNAs can act as "denoising" mechanisms to buffer fluctuations in gene expression, which warrants further experimental exploration.
Summary updated