When aspirin is used in combination with antidiabetic drugs, is there a risk of drug interactions? Intelligent systems may hold the key to resolving such critical challenges related to combined medication that directly impact patients' medication safety in the future.
Recently, Professor Zhang Wen's team from the College of Informatics at Huazhong Agricultural University (HZAU) has released an innovative breakthrough at the international conference, The 63rd Annual Meeting of the Association for Computational Linguistics. The team introduced a pairwise knowledge-augmented language model PKAG-DDI, which is capable of producing full reports on drug interactions, so as to offer intelligent support for safe medication practices.

The pairwise knowledge-augmented language model PKAG-DDI. [Photo/news.hzau.edu.cn]
In the backdrop of global aging, combination therapy emerges as a pivotal clinical strategy, yet the perils of adverse drug interactions loom large. Detecting potential events early holds significant value in enhancing medication safety. To address this, the team developed the PKAG-DDI model, not only crafting easily digestible descriptions of drug interactions but also presenting crucial biological functional insights, showcasing inference pathways, and bolstering user confidence. Moreover, the team refined the artificial intelligence used for retrieval-enhanced generation, enabling the model to autonomously retrieve the biological functions of paired drugs and integrate them into extensive language models, thereby further elevating prediction accuracy.
Comparative experiments demonstrate that, compared to prevailing classification-based and generation-based methodologies, PKAG-DDI exhibits substantial advantages across two specialized datasets and showcases robust practicality and generalization prowess in demanding scenarios. Case studies underscore its precision in forecasting a drug's biological functions and articulating exact event descriptions.