Dual-Network Cross-Learning for Metabolite-Disease Association Prediction

Area of Science:

• Biomedical Informatics
• Computational Biology
• Metabolomics

Background:

• Metabolites are increasingly linked to human diseases, offering potential for diagnosis and treatment.
• Current deep learning methods for metabolite-disease association prediction face challenges in feature integration and graph enrichment.

Purpose of the Study:

• To develop an advanced deep learning model, DCMDA, for accurate prediction of metabolite-disease associations.
• To enhance the integration of metabolic and disease features for improved predictive performance.

Main Methods:

• Constructed a heterogeneous network by integrating similarity and association networks.
• Employed non-negative matrix factorization and graph autoencoders for feature extraction.
• Utilized a Cross-Attention mechanism for feature fusion to obtain deep metabolite and disease representations.

Main Results:

• DCMDA achieved high predictive accuracy in 5-fold cross-validation.
• Demonstrated superior performance compared to existing state-of-the-art prediction methods.
• Achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 97.8% and an Area Under the Precision-Recall Curve (AUPR) of 97.9%.

Conclusions:

• DCMDA effectively addresses limitations in current metabolite-disease association prediction models.
• The proposed Dual-network Cross-learning approach offers a robust framework for uncovering complex metabolite-disease relationships.
• This model holds significant potential for advancing disease diagnosis and therapeutic strategies.