Advances and Challenges in Drug-drug Interaction Prediction

Abstract

Drug interaction (DDI) occurs when multiple drugs are used at the same time. Accurate prediction of the specific mechanisms behind DDI, known as DDI events or DDIE, is essential for clinical safe drug use. In the modern era, drug-drug drug interaction prediction is crucial to drug safety. Traditional DDI prediction has many shortcomings and deficiencies, such as DDI data is relatively limited, the generalization ability of traditional methods is weak and so on. Due to the excessive reliance of the model on the existing limited data, its prediction performance degrades dramatically when faced with brand-new drug molecules or unknown drug combinations. This makes it difficult for them to find novel ddis, which are urgently needed for drug development and clinical safety monitoring. In the latest research, people are developing towards the combination of multi-source data fusion and graph neural network. This paper aims to discuss and summarize the latest research progress in the field of DDI prediction. The research trend mentioned in this paper is changing from single data source to multi-source heterogeneous data fusion and graph neural network. By integrating the multi-modal data of drugs and using the adaptive fusion mechanism, the new model can understand the properties of drugs more comprehensively and accurately. At the same time, these innovative methods are committed to extending the prediction task from simple binary classification to refined event type prediction, thus laying the foundation for the development of more clinically practical DDI prediction models.