A structural deep network embedding model for predicting associations between miRNA and disease based on molecular...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• MicroRNAs (miRNAs) are crucial in human biological processes and diseases.
• Experimental verification of miRNA-disease associations is limited due to biotechnological constraints.
• Developing precise computational methods for predicting miRNA-disease associations is essential.

Purpose of the Study:

• To develop a novel computational model, Structural Deep Network Embedding for miRNA-Disease Association (SDNE-MDA), for predicting potential miRNA-disease associations.
• To leverage molecular association networks and integrate multi-modal data for enhanced prediction accuracy.

Main Methods:

• Integrated miRNA attribute information using the Chao Game Representation (CGR) algorithm and disease attribute information via semantic similarity.
• Extracted features from a heterogeneous molecular association network using structural deep network embedding.
• Constructed comprehensive feature descriptors by combining attribute and behavioral information.
• Utilized Convolutional Neural Network (CNN) for training and classification of feature descriptors.

Main Results:

• Achieved an Area Under the Curve (AUC) of 0.9447 and a prediction accuracy of 87.38% in five-fold cross-validation on the HMDD v3.0 dataset.
• Demonstrated superior performance compared to alternative feature extraction and classification models.
• Case studies confirmed 47, 46, and 46 top-50 predicted miRNAs for Breast Neoplasms, Kidney Neoplasms, and Lymphoma, respectively, in independent databases.

Conclusions:

• SDNE-MDA is a reliable computational tool for predicting potential miRNA-disease associations.
• The model's ability to integrate diverse data types and extract complex network features enhances prediction accuracy.
• This approach facilitates the discovery of novel miRNA-disease relationships, aiding further biological investigation.