Predicting microRNA-Disease Associations Through Multi-View Feature Fusion and Matrix Completion on Graph...

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• MicroRNAs (miRNAs) play crucial regulatory roles in complex human diseases.
• Accurate prediction of miRNA-disease associations is vital for biological studies and understanding disease mechanisms.
• Existing models for predicting miRNA-disease associations require enhancement for improved accuracy and biological insight.

Purpose of the Study:

• To propose a novel graph convolutional network-based model, FMCGCN, for predicting miRNA-disease associations.
• To leverage multi-view feature fusion and matrix completion for enhanced prediction accuracy.
• To validate the model's performance against existing methods and through case studies.

Main Methods:

• Utilized graph autoencoders to learn multiple embeddings from diverse networks.
• Employed attention mechanisms for effective fusion of multi-view features.
• Integrated graph convolutional networks (GCNs) for local information aggregation.
• Applied feature and nuclear norm minimization for matrix completion to generate prediction matrices.

Main Results:

• The proposed FMCGCN model demonstrated superior performance across several metrics compared to current state-of-the-art models.
• 5-fold cross-validation confirmed the robustness and accuracy of FMCGCN.
• Case studies on esophageal and pancreatic neoplasms validated the model's practical applicability and biological relevance.

Conclusions:

• FMCGCN offers a powerful and accurate approach for predicting miRNA-disease associations.
• The model's ability to fuse multi-view features and employ matrix completion enhances predictive capabilities.
• FMCGCN provides a valuable tool for advancing research in miRNA-mediated diseases and their potential therapeutic targets.