Predicting ncRNA-Protein interactions with a graph attention model exploiting personalized subgraphs

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Accurate prediction of non-coding RNA (ncRNA) and protein interactions is essential for gene regulation, disease mechanism elucidation, and therapeutic development.
• Machine learning, especially graph neural networks (GNNs), has shown promise in predicting these interactions by analyzing molecular data structures.
• Existing GNN methods often use fixed-hop subgraphs, which may limit their ability to capture complex interaction patterns and omit relevant molecular information.

Purpose of the Study:

• To develop a novel computational method for enhanced prediction of ncRNA-protein interactions.
• To overcome the limitations of fixed-hop subgraph analysis in current GNN approaches.
• To improve the accuracy and scalability of ncRNA-protein interaction prediction by integrating sequence and structural features.

Main Methods:

• A personalized subgraph selection framework is employed to extract the most informative subgraphs around potential interaction sites.
• A graph attention network (GAT) is utilized to learn node representations from these personalized subgraphs.
• Sequence-level features are captured using k-mer frequencies, while structural information is incorporated via node2vec embeddings.

Main Results:

• The proposed method demonstrates significant improvements in predicting ncRNA-protein interactions compared to existing approaches.
• The algorithm maintains acceptable computational complexity, making it suitable for large-scale datasets.
• Integration of sequence and structural features through personalized subgraphs enhances prediction accuracy.

Conclusions:

• The novel approach using personalized subgraphs and GAT offers a more accurate and scalable solution for ncRNA-protein interaction prediction.
• This method effectively captures diverse interaction patterns by moving beyond fixed-hop subgraph limitations.
• The findings contribute to advancing research in gene regulation, disease mechanisms, and drug discovery.