ESGC-MDA: Identifying miRNA-Disease Associations Using Enhanced Simple Graph Convolutional Networks.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• MicroRNAs (miRNAs) are crucial in human disease development.
• Accurate identification of miRNA-disease associations aids disease diagnosis and treatment.
• Efficient computational prediction methods are needed to reduce experimental costs and time.

Purpose of the Study:

• To develop an efficient computational method for predicting miRNA-disease associations.
• To address the challenge of high-quality feature representation in graph neural network-based prediction.
• To propose the Enhanced Simple Graph Convolution Network for miRNA-disease association (ESGC-MDA) prediction.

Main Methods:

• Constructed a bipartite attributed graph using multi-source similarity for miRNAs and diseases.
• Enhanced node feature representations via random message dropping and adaptive layer aggregation within a simple convolution network.
• Utilized a fully connected neural network decoder to compute miRNA-disease pair prediction scores.

Main Results:

• ESGC-MDA demonstrated superior performance compared to state-of-the-art methods in 5-fold and 10-fold cross-validation on HDMM v2.0 and HMDD v3.2 datasets.
• Case studies on cardiovascular disease, lung cancer, and colon cancer validated the model's effectiveness.
• Achieved high-quality feature representation for improved miRNA-disease association prediction.

Conclusions:

• ESGC-MDA offers an effective and efficient approach for predicting miRNA-disease associations.
• The method enhances feature representation learning in graph neural networks for this task.
• The findings support the utility of ESGC-MDA in advancing disease diagnosis and therapeutic strategies.