Related Experiment Video
Updated: Dec 6, 2025
07:35
A Knowledge Graph Approach to Elucidate the Role of Organellar Pathways in Disease via Biomedical Reports
Published on: October 13, 2023
2.0K
Deep-DRM: a computational method for identifying disease-related metabolites based on graph deep learning approaches
Tianyi Zhao1, Yang Hu2, Liang Cheng3
1Department of Computer Science at the Harbin Institute of Technology.
Briefings in Bioinformatics
|October 13, 2020
Summary
Deep-DRM, a novel graph deep learning method, effectively identifies disease-related metabolites by integrating chemical structures and biological networks. This approach significantly improves the prediction of metabolite-disease pairs (MDPs), advancing our understanding of metabolic disease mechanisms.
Area of Science:
- Metabolomics
- Systems Biology
- Bioinformatics
Background:
- Metabolic changes are crucial indicators of disease, yet understanding disease-related metabolites lags behind genes, RNAs, and proteins.
- Existing methods for identifying disease-related metabolites often overlook crucial chemical structure information and fail to capture complex association patterns.
Purpose of the Study:
- To develop an advanced computational method for identifying disease-related metabolites.
- To overcome limitations of current approaches by incorporating metabolite chemical structures and biological network data.
Main Methods:
- A graph deep learning framework, Deep-DRM, was developed.
- Metabolite similarities were derived from chemical structures; disease similarities were based on functional gene networks and semantic associations.
- Graph Convolutional Networks (GCN) and Principal Component Analysis (PCA) were employed for feature extraction and dimensionality reduction, followed by a deep neural network for prediction.
Main Results:
- Deep-DRM achieved high performance with an Area Under the Curve (AUC) of 0.952 and Area Under the Precision-Recall curve (AUPR) of 0.939 in 10-cross validations.
- The method demonstrated superior performance compared to existing approaches.
- A significant portion (10 out of 15) of top predicted disease-metabolite associations were validated by external studies.
Conclusions:
- Deep-DRM is a highly efficient and accurate method for identifying disease-related metabolites and their associations.
- The integration of chemical structure and network information provides a powerful strategy for advancing metabolomic research in disease.
- This method offers a valuable tool for discovering novel biomarkers and therapeutic targets in various diseases.
