MRDDA: a multi-relational graph neural network for drug-disease association prediction.

backgroundDrug repositioning offers a promising avenue for accelerating drug development and reducing costs. Recently, computational repositioning approaches have gained attraction for identifying potential drug-disease associations (DDAs). Biological entities such as drugs, genes, proteins, RNA, and diseases interact within a complex network. How to adequately extract the intrinsic relationships among them and accurately predict the drug-disease associations remains a challenge.

methodsIn this study, we introduce MRDDA, a novel graph neural network model that utilizes multiple relations, for drug repositioning. First, we design a hybrid graph convolutional framework to capture both local and global representations of drugs and diseases. Subsequently, a meta-path-based approach is employed to capture high-order topological representations from these entities. Finally, we present a layer-wise attention mechanism to integrate embeddings from various layers.

resultsThe MRDDA model demonstrated superior performance in predicting drug-disease associations compared to existing methods, achieving higher results during the 10-fold cross-validation on three benchmark datasets. Notably, in our case studies focusing on Alzheimer's disease and breast cancer, MRDDA effectively identified several promising drug candidates that were previously unrecognized for these conditions. Additionally, molecular docking experiments reinforced our results by confirming the binding affinities and interactions between selected drugs and their target diseases, suggesting a solid basis for further experimental studies.

conclusionMRDDA offers an innovative framework for drug repositioning by effectively modeling and predicting drug-disease associations through advanced graph neural network techniques. The model's ability to integrate multi-relational data with greater accuracy paves the way for more efficient identification of potential therapeutic uses for existing drugs, ultimately contributing to the acceleration of drug development and reduced costs in the pharmaceutical industry.