OGNNMDA: a computational model for microbe-drug association prediction based on ordered message-passing graph neural...

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning

Background:

• Existing computational models for microbe-drug association prediction have limitations.
• There is a need for improved accuracy and robustness in these predictions.

Purpose of the Study:

• To propose the OGNNMDA framework for enhanced microbe-drug association prediction.
• To leverage an ordered message-passing mechanism for improved feature representation.

Main Methods:

• Constructed a microbe-drug heterogeneous matrix by integrating multiple similarity matrices.
• Employed a multi-layer ordered message-passing graph neural network encoder for feature extraction.
• Utilized a bilinear decoder for final prediction of microbe-drug associations.

Main Results:

• OGNNMDA demonstrated superior prediction performance on the aBiofilm and MDAD datasets.
• The method achieved sub-optimal results on the DrugVirus dataset.
• Case studies validated the effectiveness of OGNNMDA in predicting known microbe-drug associations.

Conclusions:

• The OGNNMDA framework offers a promising approach for accurate microbe-drug association prediction.
• The ordered message-passing mechanism is key to its enhanced embedding capabilities.
• Further validation on diverse datasets can solidify its applicability.