Predicting miRNA-based disease-disease relationships through network diffusion on multi-omics biological data.

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are crucial regulators of gene expression implicated in complex diseases.
• Experimental identification of disease-associated miRNAs is resource-intensive.
• Computational prediction of miRNA-disease associations offers a promising avenue for therapeutic discovery and understanding disease relationships.

Purpose of the Study:

• To develop and validate an in-silico method, MiRNA-disease Association Prediction (MAP), for predicting and prioritizing miRNA-disease associations.
• To leverage network diffusion on a heterogeneous network for enhanced prediction accuracy.
• To explore the utility of MAP in identifying differential miRNA expression and uncovering disease subtypes.

Main Methods:

• Constructed a heterogeneous network integrating miRNA-gene, protein-protein, and gene-disease associations.
• Applied a network diffusion approach starting from known disease genes.
• Validated the MAP method against experimental miRNA-disease association data.

Main Results:

• MAP demonstrated superior performance compared to state-of-the-art methods, achieving areas under the ROC curve > 0.8 for four cancer types.
• Successfully predicted differential miRNA expression in four distinct cancer types.
• Identified disease-disease relationships via shared miRNAs, revealing hidden disease subtypes comparable to gene-based methods.

Conclusions:

• The MAP method provides an efficient and accurate in-silico approach for predicting miRNA-disease associations.
• MAP facilitates the identification of potential miRNA biomarkers and therapeutic targets for complex diseases.
• The approach offers novel insights into disease heterogeneity and relationships through miRNA-based subtyping.