Prediction of potential disease-associated microRNAs using structural perturbation method

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Identifying microRNA (miRNA)-disease associations is crucial but challenging.
• Existing link prediction methods focus on accuracy but neglect network predictability.
• Current approaches often overlook the inherent predictability of miRNA-disease association networks.

Purpose of the Study:

• To develop a novel computational method for predicting miRNA-disease associations.
• To evaluate the link predictability of integrated miRNA and disease networks.
• To improve the accuracy of identifying potential miRNA-disease relationships.

Main Methods:

• Constructed a bilayer network integrating miRNA-disease, miRNA similarity, and disease similarity networks.
• Utilized structural consistency as an indicator for network link predictability.
• Applied the structural perturbation method (SPM) for predicting miRNA-disease associations.

Main Results:

• The bilayer network exhibits higher link predictability than the miRNA-disease network alone.
• SPM demonstrated reliable performance in 5-fold cross-validation, outperforming other algorithms.
• A case study on breast cancer identified 80% of top-predicted miRNAs as experimentally confirmed.

Conclusions:

• The structural perturbation method (SPM) is a reliable computational tool for enhancing miRNA-disease association identification.
• Integrating multiple networks improves the predictability and accuracy of miRNA-disease association prediction.
• This approach offers a valuable method for advancing bioinformatics research in miRNA-disease association discovery.