Inferring miRNA-disease associations using collaborative filtering and resource allocation on a tripartite graph

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• MicroRNAs (miRNAs) play vital roles in biological processes.
• Experimental verification of miRNA-disease associations is costly and time-consuming.
• Efficient computational methods are urgently needed to predict these associations.

Purpose of the Study:

• To develop a novel computational method for inferring miRNA-disease associations.
• To address challenges of imbalanced data and predict associations among multiple entities.
• To provide a powerful tool for miRNA-disease association discovery.

Main Methods:

• Utilized a miRNA-disease-lncRNA tripartite graph.
• Employed collaborative filtering (CFNBC model) for imbalanced data.
• Applied resource allocation algorithms and TPGLDA model for association prediction.

Main Results:

• Achieved high performance with AUC of 0.9788 and AUPR of 0.9373.
• Outperformed existing methods like DCSMDA and TPGLDA.
• Identified numerous novel, literature-confirmed miRNA-disease associations and discovered associations for new diseases/miRNAs.

Conclusions:

• The proposed method demonstrates reliable performance in inferring miRNA-disease associations.
• It effectively discovers novel associations, including for entities with no prior known links.
• The method serves as a valuable tool for advancing miRNA-disease association research.