GACNNMDA: a computational model for predicting potential human microbe-drug associations based on graph attention...

Area of Science:

• Microbiology
• Pharmacology
• Computational Biology

Background:

• Human microbes are increasingly recognized as crucial drug targets linked to health.
• Predicting microbe-drug associations is vital for drug research but challenging due to limited known interactions.
• Existing methods struggle with predicting interactions for novel microbes or drugs.

Purpose of the Study:

• To develop an effective computational model for predicting microbe-drug associations.
• To address the challenge of limited known microbe-drug interaction data.
• To facilitate drug discovery and development by identifying potential microbe-drug pairs.

Main Methods:

• Constructed two heterogeneous microbe-drug networks using similarity measures and known associations.
• Developed feature matrices for microbes and drugs by concatenating their attributes.
• Employed a two-layer graph attention network (GAT) to obtain low-dimensional feature representations.
• Integrated GAT outputs with feature matrices into a convolutional neural network (CNN) to create the GACNNMDA model.

Main Results:

• The GACNNMDA model demonstrated superior predictive performance compared to existing advanced methods.
• Experimental results confirmed the effectiveness of GACNNMDA in predicting microbe-drug associations.
• Case studies validated the model's capability using well-known microbes and drugs.

Conclusions:

• GACNNMDA offers a powerful computational approach for predicting microbe-drug associations.
• The model's accuracy and effectiveness can significantly aid drug research and development.
• This work provides a valuable tool for exploring the complex relationship between microbes and drugs.