Predicting Combinative Drug Pairs via Integrating Heterogeneous Features for Both Known and New Drugs

摘要

An ordinary disease caused by the anomaly of expression level of an individual gene, can be treated by a specific drug, which regulates the gene's expression. However, this single-drug treatment has very low effectiveness on complex diseases [1], which usually involve multiple genes and pathways of the metabolic network. Drug combination, as one of multiple-target treatments, has demonstrated its effectiveness in treating complex diseases, such as HIV/AIDS [2] and colorectal cancer [3]. However, it is still costly and time-consuming in clinical trials to find an effective combination of individual drugs. Fortunately, both the number of approved drug combinations [4] and the amount of available heterogeneous information about drugs are increasing. It became feasible to develop computational approaches to predict potential candidates of drug pairs for the treatments of complex diseases [5, 6].