Prediction of Human Microbe-Drug Association based on Layer Attention Graph Convolutional Network.

Area of Science:

• Microbiology and Pharmacology
• Computational Biology and Bioinformatics
• Genomics and Drug Discovery

Background:

• Human microbes are linked to complex diseases and are emerging as critical drug targets.
• Identifying microbe-drug associations is vital for drug development, precision medicine, and addressing antimicrobial resistance.

Purpose of the Study:

• To propose a novel computational model for predicting microbe-drug associations.
• To leverage a layer attention graph convolutional network for enhanced prediction accuracy.

Main Methods:

• Integrated multiple biological data into a heterogeneous network.
• Employed a graph convolutional network (GCN) to embed microbes and drugs.
• Utilized a layer attention mechanism to decode embeddings and calculate association scores.

Main Results:

• Achieved high Area Under the Curve (AUC) values, e.g., 0.9178 on aBiofilm dataset using global LOOCV.
• Demonstrated strong performance in 5-fold cross-validation with average AUCs of 0.9141 and 0.8982 for aBiofilm and MDAD datasets, respectively.
• Case studies further validated the model's predictive capabilities.

Conclusions:

• The proposed computational model offers an efficient alternative to time-consuming traditional methods for microbe-drug association prediction.
• The model shows significant potential for advancing drug development and precision medicine.
• Accurate prediction of microbe-drug associations can aid in developing new therapies and combating drug resistance.