DrugBank-2025: An AI-Ready Next-Generation Dataset for Drug–Drug Interaction Prediction

Accurate prediction of Drug–Drug Interactions (DDIs) is critical for reducing adverse drug reactions and ensuring safe combination therapies. However, existing benchmarks for DDI prediction suffer from outdated datasets and limited scope, hindering comprehensive evaluation of emerging AI methodologies. To address this gap, we present DrugBank-2025, a large-scale, diverse, and up-to-date dataset curated from the latest DrugBank release. DrugBank-2025 is designed as an AI-ready benchmark resource that supports systematic evaluation across various DDI prediction tasks, including binary, multi-class, and multi-label classification. By providing extensive coverage of approved drugs and interaction types, DrugBank-2025 enables robust assessment of model generalization under realistic and challenging scenarios, such as out-of-distribution drug pairs, facilitating the development and benchmarking of next-generation DDI prediction models.