Understanding and preventing drug side effects holds a profound influence on drug development and utilization, profoundly impacting patients’ physical and mental well-being. Traditional artificial drug experimentation methods are not only expensive but also time-consuming, rendering comprehensive testing a challenging task. However, with the advent of advanced technologies, particularly in the realm of machine learning and the availability of extensive biochemical medical data, combining these two has emerged as a pivotal approach for predicting drug side effects. Recently, the expanding literature on drug side effects, coupled with the proliferation of websites and databases containing comprehensive information on drugs, side effects, and associated biological entities, offers researchers ample opportunities to gather data and advance the development of machine learning methods for predicting drug side effects. While numerous review articles have addressed drug side effect prediction, most have concentrated on methods for predicting drug side effect associations.

To solve the problems, a research team led by Haochen Zhao published their new review on 15 May 2025 in Frontiers of Computer Science co-published by Higher Education Press and Springer Nature.

They sorts out machine learning-based prediction methods for side effects caused by single drugs and DDIs. Moreover, the study of prediction of the frequency and severity of side effects for drugs is highlighted. They outline the fundamental principles behind establishing predictive models and introduce commonly utilized databases and web servers employed in the detection of drug side effects. Finally, they deliberate on the current challenges and future avenues in machine learning-based methods for discerning drug side effects.

DOI: 10.1007/s11704-024-31063-0