Recent Developments in Link Prediction for Drug Interaction Networks

Drug-drug interactions also known as DDIs are dangerous phenomena in the sphere of pharmacology and healthcare, contributing to negative side effects and problems in the treatment of patients. Predicting the likely DDIs is an important effort in drug development and application, individualized therapeutics and the overall patient care and safety. In this research, link prediction approach from network science and machine learning are applied to discover the unseen relationships between drugs. We use drugs as nodes and interactions as edges of the graph as well as use different predictive algorithms which may include similarity measures, machine learning classification, and deep learning to predict new edges between drugs that have not interacted in the past. To increase the level of prediction accuracy, we consider a wide range of characteristics that are based on chemical structures, biological pathways, and pharmacological actions. The outcome of our suggestive frame work is measure on standard data sets-which show its usefulness for predicting both known and novel drug interactions. Here, we have compared the performance of DDI predictions based on both multi-dimensional data and link prediction and it has been shown that the link prediction models enhance the reliability of the predictions. This study seems to offer a systematic way to identify potentially problematic drug-drug interactions therefore could eventually help safeguard patient from adverse effects resulting from drug prescription and potentially aid drug development process.