Deep learning in stroke therapeutics: drug repurposing and beyond

Deep learning is reshaping stroke research by accelerating drug repurposing amid heterogeneous pathology, narrow therapeutic windows, and poor translation. This review highlights current therapeutic challenges and emerging DL applications from preclinical modeling to clinical decision support.

This narrative review focuses on the application of DL in preclinical and clinical stroke research, with particular emphasis on their roles in drug discovery and repurposing, as well as the current limitations of these approaches. PubMed was searched for peer-reviewed studies using keywords related to drug repurposing, stroke, and computational approaches published between 2020 and 2025.

Given the global burden of stroke and limited therapeutic options, DL offers a timely solution by enabling accelerated drug repurposing and efficient drug development. Its ability to analyze high-dimensional data contributes to target identification, virtual screening, and drug repurposing that bridges translational gaps in stroke research. The approval of multiple AI-based diagnostic tools by regulatory bodies like the US FDA reflects growing clinical adoption. However, challenges remain in model interpretability, generalizability, and real-world validation.