Predicting miRNA-Disease Associations via Meta-Path Embedding.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• MicroRNAs (miRNAs) are crucial regulators in biological processes.
• Accurate identification of miRNA-disease associations is vital for disease research.
• Existing prediction methods often lack interpretability.

Purpose of the Study:

• To develop a novel, interpretable computational method for predicting miRNA-disease associations.
• To leverage network embedding techniques for enhanced feature extraction.
• To improve the accuracy and reliability of miRNA-disease association predictions.

Main Methods:

• Utilized metapath-based network embedding (metapath2vec) on a heterogeneous miRNA-disease interaction network.
• Employed the XGBoost classifier for predicting potential miRNA-disease associations.
• Validated the method's performance using AUC, AUPR, case studies, and survival analysis.

Main Results:

• MNEMDA achieved high performance with an overall AUC of 0.959 and AUPR of 0.952.
• Outperformed state-of-the-art models with an average AUC of 0.958 across 15 diseases.
• Case studies demonstrated high validation rates (93% for renal cell carcinoma, 97% for breast neoplasms).
• Showcased robustness, generalization ability, and insensitivity to data noise.

Conclusions:

• MNEMDA is a robust and interpretable tool for predicting miRNA-disease associations.
• The method offers significant improvements over existing approaches.
• MNEMDA holds practical value for disease research and potential therapeutic target identification.