EPMDA: Edge Perturbation Based Method for miRNA-Disease Association Prediction

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are implicated in numerous human diseases.
• Accurate prediction of miRNA-disease associations is crucial for understanding disease mechanisms and developing treatments.

Purpose of the Study:

• To propose a novel edge perturbation-based method (EPMDA) for predicting potential miRNA-disease associations.
• To enhance the accuracy of identifying relationships between miRNAs and diseases.

Main Methods:

• Developed a feature vector using structural Hamiltonian information to describe graph edges.
• Employed a multi-layer perception model trained on extracted features for association prediction.
• Utilized an edge perturbation approach for enhanced prediction.

Main Results:

• EPMDA achieved an AUC of 0.9818 in 5-fold cross-validation on the HMDD dataset, outperforming DeepMDA by 3.5%.
• Achieved an AUC of 0.9371 in leave-one-disease-out cross-validation, surpassing DeepMDA by 7.4%.
• Case studies confirmed 42, 46, and 41 top-50 predicted miRNAs for three diseases, validating EPMDA's performance.

Conclusions:

• EPMDA demonstrates superior performance in predicting miRNA-disease associations compared to existing methods.
• The proposed method offers a valuable tool for advancing research in miRNA-related diseases.
• Findings support the utility of structural graph information and machine learning in bioinformatics.