GSAMDA: a computational model for predicting potential microbe-drug associations based on graph attention network and...

Area of Science:

• Microbiology
• Pharmacology
• Computational Biology

Background:

• Microorganisms significantly impact human health.
• Discovering microbe-drug associations is crucial for drug research and development.
• Limited computational methods exist for predicting these associations.

Purpose of the Study:

• To propose a novel computational model for inferring latent microbe-drug associations.
• To develop an effective tool for advancing drug discovery through microbial insights.

Main Methods:

• Developed GSAMDA, a model integrating graph attention networks (GAT) and sparse autoencoders.
• Constructed a heterogeneous network with known microbe-drug associations and similarities.
• Employed GAT and sparse autoencoder modules to learn topological and attribute representations.

Main Results:

• GSAMDA effectively infers potential microbe-drug associations.
• The model achieved superior performance compared to five other state-of-the-art methods.
• Case studies validated the model's effectiveness for specific drug and microbe categories.

Conclusions:

• GSAMDA offers a powerful new approach for predicting microbe-drug associations.
• The model shows significant potential for facilitating drug research and development.
• Experimental results and case studies confirm the model's high performance and utility.