Contrastive hypergraph collaborative filtering for transfer RNA-disease association prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Transfer RNAs (tRNAs) are vital for protein synthesis and cellular homeostasis.
• Emerging evidence links tRNAs to various disease progressions.
• Accurate prediction of tRNA-disease associations is crucial for disease mechanism research and precision medicine.

Purpose of the Study:

• To develop an advanced computational framework for predicting tRNA-disease associations.
• To overcome limitations of existing methods in handling complex and heterogeneous data.
• To enhance the accuracy and reliability of tRNA-disease association predictions.

Main Methods:

• Introduced Contrastive Hypergraph Collaborative Filtering (CoHGCL), integrating hypergraph contrastive learning and collaborative filtering.
• Employed Graph Attention Networks and Random Walk with Restart for structural and topological feature extraction.
• Utilized node-level contrastive learning for multiview feature embedding and generalized matrix factorization/MLPs for association modeling.

Main Results:

• CoHGCL demonstrated superior performance in five-fold cross-validation compared to existing methods.
• Achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.9623 and Area Under the Precision-Recall Curve (AUPRC) of 0.9430.
• Case studies validated CoHGCL's capability in discovering novel and biologically relevant tRNA-disease associations.

Conclusions:

• CoHGCL offers a powerful and effective approach for predicting tRNA-disease associations.
• The framework advances the understanding of tRNA roles in disease pathogenesis.
• CoHGCL provides a valuable tool for precision medicine and future biomedical research.