Predicting Drug–Gene–Disease Associations by Tensor Decomposition for Network-Based Computational Drug Repositioning

Drug repositioning offers the significant advantage of greatly reducing cost and time drug discovery by identifying new therapeutic indications for existing drugs. In particular, computational approaches using networks in have attracted attention inferring potential associations between drugs diseases efficiently based on network connectivity. this article, we proposed a network-based method to construct drug–gene–disease tensor integrating drug–disease, drug–gene, disease–gene predict triple through decomposition. The method, which ensembles generalized decomposition (GTD) multi-layer perceptron (MLP), models GTD learns features drugs, genes, MLP, providing more flexibility non-linearity than conventional We experimented with association prediction two distinct created chemical structures ATC codes as features. Moreover, leveraged drug, gene, disease latent vectors obtained from predicted pairwise associations. Our experimental results revealed that ensemble was superior prediction. model achieved an AUC 0.96 predicting resulting approximately 7% improvement over performance models. It also showed competitive accuracy compared previous methods. This study demonstrated incorporating genetic information leads notable advancements repositioning.