Identification of metabolite-disease associations based on knowledge graph

Area of Science:

• Biomedical Informatics
• Computational Biology
• Metabolomics

Background:

• Traditional wet lab experiments for metabolite analysis are time-consuming and labor-intensive.
• Metabolite analysis offers critical insights into disease onset, progression, and potential treatments.
• Developing efficient methods for identifying disease-metabolite associations is crucial for advancing medical research.

Purpose of the Study:

• To develop a novel machine learning model, COM-RAN, for identifying potential associations between metabolites and diseases.
• To overcome the limitations of traditional experimental methods by leveraging computational approaches.
• To enhance the accuracy and efficiency of disease-metabolite association prediction.

Main Methods:

• Integration of known disease-metabolite associations.
• Synthesis of existing data on diseases and metabolites with supplementary information.
• Characterization of disease-metabolite associations using knowledge graph-based embedded features.
• Construction of a predictive model using a random forest algorithm.

Main Results:

• The COM-RAN model achieved a high Area Under the Receiver Operating Characteristic Curve (AUC) of 0.968 and an Area Under the Precision-Recall Curve (AUPR) of 0.901 in 5-fold cross-validations.
• The model outperformed most existing prediction methods in identifying disease-metabolite associations.
• Case studies validated a majority of the novel associations predicted by COM-RAN, confirming its reliability.

Conclusions:

• The COM-RAN model shows significant promise for predicting disease-metabolite associations.
• Integrating knowledge graphs with machine learning enhances the accuracy and reliability of predictions.
• This approach offers a powerful tool for discovering novel disease-associated metabolites and advancing personalized medicine.