Identifying Associations Between Small Nucleolar RNAs and Diseases via Graph Convolutional Network and Attention...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Small nucleolar RNAs (snoRNAs) are critical in biological processes, but their disease associations are not fully understood.
• Identifying snoRNA-disease links is vital for disease pathogenesis research.
• Conventional experimental methods for association discovery are time-consuming and expensive.

Purpose of the Study:

• To develop an efficient computational method for identifying potential small nucleolar RNA-disease associations.
• To leverage graph convolutional networks and multi-view graph attention mechanisms for this task.

Main Methods:

• Calculated snoRNA and disease similarity matrices using biological entity information.
• Employed a random forest model to weight edges between snoRNA and disease nodes.
• Constructed homogeneous and heterogeneous graphs, extracting features via graph convolutional networks and integrating them using a multi-view graph attention mechanism.
• Utilized a multilayer perceptron neural network with global and interaction features for final association prediction.

Main Results:

• The proposed GCASDA method achieved high performance with AUC of 0.9356 and AUPR of 0.9294.
• GCASDA significantly outperformed existing state-of-the-art methods across various evaluation metrics.
• A case study validated the practical feasibility and effectiveness of the GCASDA approach.

Conclusions:

• GCASDA provides an efficient and accurate computational approach for discovering novel snoRNA-disease associations.
• This method can accelerate research into disease pathogenesis by identifying key snoRNA roles.
• The findings highlight the potential of graph-based deep learning models in biological association discovery.