ZNGEA: ZINB-NMF Integrated Graph Embedding Autoencoder for Metabolite-Disease Association Identification

Area of Science:

• Computational biology
• Bioinformatics
• Systems biology

Background:

• Metabolism is crucial for life, and altered metabolites are linked to diseases.
• Traditional experimental methods for identifying metabolite-disease links are time-consuming and labor-intensive.
• Computational approaches are needed for efficient identification of metabolite-disease associations.

Purpose of the Study:

• To develop a novel deep learning algorithm, ZNGEA, for predicting potential associations between metabolites and diseases.
• To overcome the limitations of traditional experimental methods in identifying metabolite-disease links.

Main Methods:

• ZNGEA integrates Zero-Inflated Negative Binomial (ZINB) distribution and Non-negative Matrix Factorization (NMF).
• Combines multiple disease and metabolite similarity networks using a nonlinear method.
• Applies NMF and a ZINB-based graph convolutional autoencoder for feature extraction.
• Utilizes a bilinear decoder for training the model.

Main Results:

• ZNGEA achieved an Area Under the Curve (AUC) of 0.9859 and Area Under the Precision-Recall Curve (AUPR) of 0.9820 in 5-fold cross-validation.
• Performance surpassed existing methods.
• Case studies validated the majority of newly identified metabolite-disease links, confirming ZNGEA's reliability.

Conclusions:

• ZNGEA is a reliable and efficient computational tool for predicting metabolite-disease associations.
• The method offers a valuable resource for biomedical research in exploring potential links between metabolites and diseases.
• Source code and datasets are publicly available for reproducibility and further research.