TL;DR
This paper introduces MISIM v2.0, an upgraded web server for predicting functional similarities between microRNAs (miRNAs) and their associations with diseases. It addresses limitations of previous methods by incorporating directional (up/down-regulated) miRNA-disease associations, enabling the prediction of novel associations, and providing integrated network visualization and functional enrichment analysis.
What Problem Does This Paper Address?
MicroRNAs (miRNAs) are small non-coding RNA molecules crucial for regulating gene expression and are implicated in various human diseases. Understanding the functional relationships between miRNAs is vital for predicting their roles in disease and identifying potential therapeutic targets. Previous computational methods, including the original MISIM, primarily focused on non-directional miRNA-disease associations and often produced non-negative similarity scores. This overlooked the fact that miRNAs can have opposing functional effects (e.g., promoting vs. inhibiting apoptosis), meaning their functional similarity could be negative. Furthermore, existing tools lacked integrated functionalities for network visualization and novel miRNA-disease association prediction, requiring users to employ separate software.
Methods
MISIM v2.0 builds upon the original MISIM algorithm by introducing several key improvements:
- Directional miRNA-Disease Associations: The core innovation is the incorporation of directional information (up-regulated or down-regulated) for miRNA-disease associations. This is achieved by modifying the calculation of disease semantic values to include a sign based on the deregulation type.
- Enhanced Functional Similarity Calculation: The functional similarity between two miRNAs is calculated using a cosine correlation of their semantic feature vectors. These vectors are derived from the directional miRNA-disease associations and the semantic similarity of the associated diseases. The method now accounts for the intersection of disease sets and improves the semantic feature representation by incorporating disease semantic similarity.
- Novel miRNA-Disease Association Prediction: MISIM v2.0 implements an algorithm to predict novel miRNA-disease associations. This prediction can be directional, indicating whether the association is positive or negative.
- Integrated Analysis Tools: The web server now includes built-in functionalities for:
- Network Visualization: Users can visualize miRNA functional networks, with options to display directional links in different colors.
- Functional Enrichment Analysis: The server integrates the TAM 2.0 algorithm for dissecting enriched functions of paired or networked miRNAs.
Datasets
- MeSH Terms: Obtained from the National Library of Medicine.
- Human miRNA-Disease Association Dataset: Sourced from the HMDD v3.0 database (http://www.cuilab.cn/hmdd). This dataset was curated using MeSH disease terms and includes 14,250 miRNA-disease associations involving 1044 miRNAs and 613 diseases. miRNA information was updated using miRBase Release 22.1.
- miRNA Sets: 547 miRNA sets were obtained from TAM 2.0 (http://www.scse.hebut.edu.cn/tam/).
Key Insights and Ideas
- Directionality Matters: Recognizing that miRNAs can have opposing functional roles (e.g., activating vs. inhibiting cellular processes) is crucial for accurately assessing their functional similarity. MISIM v2.0's ability to capture negative functional similarity is a significant step forward.
- Integrated Platform Value: Consolidating data, similarity calculation, network visualization, and prediction into a single web server greatly enhances usability and facilitates downstream analysis for researchers.
- Predictive Power: The ability to predict novel, and importantly, directional miRNA-disease associations opens new avenues for hypothesis generation and experimental validation.
Key Results
- Data Expansion: MISIM v2.0 contains 1044 miRNAs and 613 diseases, a 3-fold increase compared to MISIM v1.0.
- Directional Similarity: The server can now produce both positive and negative functional similarity scores between miRNAs, reflecting their potentially opposing roles.
- Novel Association Prediction: The server can predict novel miRNA-disease associations, including their directional nature. For example, the paper mentions predicting that hsa-mir-29 is negatively associated with Mental Disorders, with a weighted frequency of -0.26.
- Network Visualization: The server provides visual representations of miRNA functional networks, allowing users to explore relationships. Figure 2 demonstrates these networks, with one showing directional links and another without.
Limitations and Future Work
- Incomplete Data: The authors acknowledge that miRNA-disease association data is incomplete, which can introduce bias into similarity calculations and subsequent predictions.
- Pre-miRNA vs. Mature miRNA: Currently, MISIM v2.0 only runs for pre-miRNAs and cannot differentiate between specific 5p and 3p mature miRNAs. Future work could involve improving HMDD data to include mature miRNAs.
- Species Specificity: The current version is limited to human miRNAs due to the HMDD database. The authors plan to expand MISIM to include data from other species (e.g., mouse, rat).
- Browser Compatibility: The authors will continue testing MISIM v2.0 across different internet browsers to ensure broader accessibility.
- Integration with Other Methods: To further improve accuracy, the authors suggest integrating other computational methods, such as those based on Gene Ontology (GO) terms or protein-protein interaction (PPI) network topologies, into the functional similarity calculation.
Summary updated