A Multi-View Graph-Based Scheme for Drug-Drug Interactions Categorization

With the increasing significance of poly-drug therapy, the research on drug-drug interaction (DDI) has garnered more attention. Predicting potential DDI is important for ensuring patient safety and optimizing drug combinations. Recently, intelligent algorithms, such as machine learning and deep learning techniques, have emerged as effective tools for predicting potential DDI. However, most existing techniques only utilize single features for learning and describe the extraction and process of DDI data roughly. In this paper, a Multi-View Graph-Based Scheme for Drug-Drug is proposed, named MVGDDI, which aims to comprehensive DDI prediction. This scheme consists of two major parts including data acquisition and graph-based analysis. Specifically, data acquisition extracts relevant drug information from multiple drug databases using untangle. Then, the data is assembled in two different ways to meet specific requirements. At last, the multiple features are organized into a multi-view dataset for comprehensive learning. Subsequently, a graph-based model is adopted to learn the relationships among DDI data and categorize DDIs. The experiments with the actual anticancer drug demonstrate the effectiveness and superiority of MVGDDI compared to the conventional machine learning methods.