Predicting miRNA-Disease Associations via Node-Level Attention Graph Auto-Encoder.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are crucial non-coding RNAs involved in biological processes and human diseases.
• Accurate prediction of miRNA-disease associations aids in understanding disease mechanisms at the molecular level.
• Traditional experimental methods for miRNA-disease association prediction are time-consuming and costly.

Purpose of the Study:

• To propose a novel computational method, AGAEMD, for predicting potential miRNA-disease associations.
• To develop an efficient and accurate approach for miRNA-disease association inference.

Main Methods:

• Constructed a heterogeneous matrix integrating miRNA similarity, disease similarity, and known miRNA-disease associations.
• Employed a node-level attention encoder-decoder network (AGAEMD) utilizing low-dimensional embeddings.
• Calculated association scores to predict potential miRNA-disease links.

Main Results:

• AGAEMD demonstrated superior performance compared to state-of-the-art methods on benchmark datasets (HMDD v2.0 and v3.2).
• Experimental results showed excellent predictive accuracy and reliability.
• Case studies on Colon Neoplasms, Lung Neoplasms, and Lupus Vulgaris confirmed AGAEMD's effectiveness.

Conclusions:

• AGAEMD provides an efficient and accurate computational tool for predicting miRNA-disease associations.
• The method offers a valuable alternative to experimental approaches, facilitating disease mechanism research.
• AGAEMD shows reliable predictive performance, aiding in the discovery of novel miRNA-disease links.