1136-Deep Learning Architectures for Drug Synergy Prediction

This review explores the evolution and application of deep learning for predicting anti-cancer drug synergy, a therapeutic approach that improves efficacy while minimizing side effects. It categorizes computational models into single-task and multi-task learning frameworks, detailing how they utilize biomedical data such as chemical structures and genomic profiles. The authors highlight various benchmarking datasets and interactive tools while comparing different architectural strategies, including branch and graph structures. Despite significant progress, the review identifies critical challenges regarding drug concentration modeling and the need for more sophisticated feature integration. Ultimately, the text proposes that optimizing multi-task architectures and designing new auxiliary tasks, such as drug-target affinity prediction, will be essential for future breakthroughs.

References:

• Li, L., Zhang, H., Zheng, C. et al. A review of deep learning approaches for drug synergy prediction in cancer. npj Drug Discov. 2, 30 (2025). doi.org