Knowledge Graph Representation Learning Based Drug Informatics
Rajeev Kumar Verma, Preetam Kumar · 2019
Drug Discovery and assessment is increasingly a complex albeit an acute process. Current approaches to the study of drugs involving clinical evaluation and post-marketing surveillance is very time consuming and costly. Though the machine-learning based systems have been proposed in the literature, but those systems are task-dependent. This work presents a rather data-intensive approach to the study of drugs using the large-scale DrugBank dataset, a Linked Open Dataset to further the process of drug discovery. We used Representation Learning approach on the large-scale drug dataset which not only provides automated framework for feature learning for machine learning task but also helps in prediction of new facts in the knowledge base through Link-Prediction. The proposed representation learning system is promising to be used general approach to solve different problems and we experiment on Drug-Drug Interaction prediction and Drug-Target Prediction with comparable with respect to other work in the former task.