Predicting miRNA-Disease Associations Based On Multi-View Variational Graph Auto-Encoder With Matrix Factorization.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are crucial regulators in biological processes and human diseases.
• Understanding miRNA-disease associations is key to deciphering complex disease mechanisms.
• Traditional experimental methods for identifying these associations are costly and time-consuming.

Purpose of the Study:

• To develop an efficient computational model for predicting miRNA-disease associations.
• To leverage integrated biological data for improved association prediction.
• To provide a novel tool for researchers in the field of miRNA-related diseases.

Main Methods:

• Developed a Variational Graph Auto-Encoder with Matrix Factorization (VGAMF) model.
• Integrated multiple miRNA and disease similarity information into comprehensive networks.
• Combined non-linear (Variational Graph Auto-Encoder) and linear (Non-negative Matrix Factorization) representations for prediction.

Main Results:

• VGAMF achieved high performance in 10-fold cross-validation, with AUCs of 0.9280 (HMDD v2.0) and 0.9470 (HMDD v3.2).
• The model outperformed existing competing methods in predicting miRNA-disease associations.
• Case studies demonstrated VGAMF's effectiveness in identifying novel associations for colon and esophageal cancers.

Conclusions:

• VGAMF is a powerful computational approach for predicting miRNA-disease associations.
• The model offers an efficient alternative to traditional experimental methods.
• This work contributes to a better understanding of miRNA roles in human diseases.