A Literature Knowledge Discovery Model Based on Graph Neural Network
Yuxiang Lang, Jingyan Huang, Ning Wu, Yanping Yang · 2024
This study developed a literature knowledge discovery model based on graph neural network and applied it in the field of potential drug discovery for Covid-19. The study utilized the Therapeutic Target Database (TTD) to construct a network of drug-target-disease relationships, and employed an association prediction method using graph convolutional neural network to discover potential connections between drugs and new coronavirus targets. A total of 5,358 “intermediary target-candidate drug” pairs were identified, revealing 1,624 potential relationships, ultimately leading to the identification of 50 potential new coronavirus treatment drugs. Subsequent literature review revealed that 28 of these drugs are currently undergoing clinical validation and have shown promising therapeutic effects against COVID-19. Additionally, 5 drugs were identified through various bioinformatics methods, while the remaining drugs have not yet been reported. The results of this research demonstrate that the literature knowledge discovery model based on graph neural networks can effectively unveil valuable insights from existing literature and facilitate the discovery of drugs with potential therapeutic benefits for COVID-19.