MicroRNA-Disease Association Prediction Models

MicroRNAs are small non-coding RNA molecules that modulate gene expression post-transcriptionally and have been implicated in the onset and progression of a wide array of human diseases. Experimental validation of microRNA–disease associations, while definitive, is laborious and costly. Computational prediction models have therefore become indispensable for prioritising candidate microRNAs for further study. These models typically integrate heterogeneous biological data—such as known association networks, sequence or functional similarities, semantic information on disease phenotypes and network topology—within frameworks ranging from kernel-based learning and graph inference to ensemble machine-learning and deep-learning approaches. Through rigorous cross-validation and case studies in cancer and other complex diseases, these methods demonstrate high predictive accuracy and robustness, guiding experimentalists towards the most promising microRNA–disease hypotheses and accelerating biomarker discovery and therapeutic development.

Research from Nature Portfolio

A network-based scoring method was introduced that quantifies both “within-disease” and “between-disease” similarity scores by integrating functional similarity of microRNAs and semantic similarity of diseases into a unified heterogeneous network. This approach is able to predict associations even for diseases with no previously known linked microRNAs and has demonstrated strong performance in leave-one-out cross-validation and in case studies across multiple cancers. Separately, a probabilistic deep-learning framework based on a restricted Boltzmann machine was developed to infer not only the presence but also the type of microRNA–disease association, thereby offering refined hypotheses about underlying molecular mechanisms. By modelling interaction patterns in a latent feature space and leveraging known association types, this model delivers reliable predictions and provides insights into diverse regulatory roles of microRNAs in pathology.

MicroRNA-Disease Association Prediction Models publication trend

The graph below shows the total number of articles in microrna-disease association prediction models across all publications each year (not limited to Nature Index journals).

Technical terms

Heterogeneous network: A graph comprising multiple types of nodes (for example, microRNAs and diseases) and diverse edges reflecting different biological relationships.

Functional similarity: A measure of likeness between two microRNAs based on shared target genes, pathways or co-expression patterns.

Semantic similarity: Quantification of resemblance between disease terms using controlled vocabulary hierarchies, such as the Disease Ontology.

Kernel similarity: A method to compute pairwise similarities by mapping entities into a high-dimensional feature space via kernel functions.

Cross-validation: A statistical technique for assessing how well a predictive model generalises to independent data by partitioning the dataset into training and testing subsets.

Area under the ROC curve (AUC): A performance metric indicating a model’s ability to distinguish between positive and negative associations, with higher values signifying better predictive power.

Restricted Boltzmann machine: A two-layer probabilistic neural network that learns latent features of input data through unsupervised training.

Matrix decomposition: A computational approach that factorises a data matrix into lower-rank components to uncover latent structure and denoise sparse observations.

References

1. WBSMDA: Within and Between Score for MiRNA-Disease Association prediction. Scientific Reports (2016).
2. RBMMMDA: predicting multiple types of disease-microRNA associations. Scientific Reports (2015).
3. HMDD v4.0: a database for experimentally supported human microRNA-disease associations. Nucleic Acids Research (2023).
4. PBMDA: A novel and effective path-based computational model for miRNA-disease association prediction. PLOS Computational Biology (2017).
5. MDHGI: Matrix Decomposition and Heterogeneous Graph Inference for miRNA-disease association prediction. PLOS Computational Biology (2018).