IGCNSDA: unraveling disease-associated snoRNAs with an interpretable graph convolutional network

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Accurate short nucleolar RNA (snoRNA)-disease association prediction is vital for disease diagnostics and therapeutics.
• Traditional experimental methods for identifying snoRNA-disease links are resource-intensive and lack scalability.
• Existing deep learning approaches often function as black boxes, limiting mechanistic understanding.

Purpose of the Study:

• To develop an interpretable deep learning model for predicting snoRNA-disease associations.
• To elucidate the underlying mechanisms connecting snoRNAs and diseases.
• To provide a scalable and efficient tool for identifying novel snoRNA-disease relationships.

Main Methods:

• Introduced IGCNSDA, an interpretable graph convolutional network (GCN) model.
• Utilized a bipartite snoRNA-disease graph to extract node feature representations.
• Developed a subgraph generation algorithm to group similar snoRNAs and diseases.
• Employed iterative embedding updates through neighbor information aggregation within subgraphs.

Main Results:

• IGCNSDA demonstrated superior performance compared to state-of-the-art methods.
• Interpretability analysis confirmed the model's ability to capture snoRNA-disease similarity.
• Case studies validated IGCNSDA's utility in predicting potential snoRNA-disease associations.

Conclusions:

• IGCNSDA offers an effective and interpretable approach for snoRNA-disease association prediction.
• The model provides valuable insights into the mechanisms underlying these associations.
• IGCNSDA serves as a powerful tool for advancing disease research and therapeutic development.