药物关联预测研究技术进展

Advances in Drug-drug Interactions Prediction Research Technology

药物关联(DDIs)是指同时或先后服用两种或两种以上药物时会导致意想不到的副作用产生。DDIs预测就是药物相互作用预测。由于实验室研究具有复杂、昂贵且耗时的缺点,因此需要发展新的预测方法来检测DDIs。DDIs预测方法可分为基于相似性、基于机器学习、基于知识图谱、基于分类、基于网络等五大类预测方法。分析了不同药物关联预测代表性方法的由来、使用场景、优缺点、作用及影响等,并阐述了DDIs预测研究所用到的数据来源。最后讨论分析了DDIs预测面临的挑战、解决方法及未来发展方向。

Drug-drug interactions(DDIs)refer to the unexpected side effects of two or more drugs taken simultaneously or sequentially.DDIs prediction is drug interaction prediction.Laboratory researches are often complex,expensive and time-consuming,which need to develop new prediction methods to detect DDIs.DDIs prediction methods can be divided into five types such as similarity-based,machine-learning-based,knowledge graph-based,classification-based and network-based.This article analyzed the origins,use scenarios,advantages and disadvantages,roles and implications of different representative methods for DDIs prediction,and expatiated the data sources used for DDIs prediction.Finally,the challenges,solutions and future development directions for DDIs prediction were discussed.