Drug2Vec: Knowledge-aware Feature-driven Method for Drug Representation Learning

Proper representations of drugs have broad applications in healthcare analytics, such as drug-drug interaction (DDI) prediction and drug-drug similarity (DDS) computation. However, drug application involves accurate drug representation and rich annotated data, requiring tremendous expert time and effort. Thereby, drug feature sparseness creates a substantial barrier for drug representation learning, making it difficult to accurately identify new drug properties prior to public release. To alleviate these deficiencies, we propose a knowledge-aware feature-driven method (Drug2Vec) for exploring the interaction between two drugs. The method of Drug2Vec captures the medical information, taxonomy information and semantic information of drugs. The results of experiments demonstrate that compared with existing methods, Drug2Vec can effectively learn the drug representation and discover accurate drug-drug interaction.