A Drug Repositioning Method Based on Heterogeneous Graph Neural Network

Automated drug repositioning can find potential drugs accurately and reduce R&D costs. To implement a drug repositioning system, first, the author builds a heterogeneous information network of drugs, diseases, and other types of nodes based on the heterogeneous network theory. Second, the meta-path model is introduced, and node and network topology information is learned by deep learning methods. And the interpretability of the model is improved by the attention mechanism. Experimental results on public data sets show that this method has reached state-of-the-art performance, and visual interpretability analysis of one of the inference cases is carried out. At the end of the article, the author provides the potential drugs for Alzheimer’s disease inferred from the model and cites relevant literature to prove its effectiveness.