Performance evaluation and application value of large language models in the prediction of drug-drug interactions

With the rapid development of artificial intelligence technology, the application of large language model (LLM) in the field of biomedical and clinical applications has gradually deepened. Drug-drug interaction (DDI) prediction is a kind of safety study with complex characteristics in drug development and use, and traditional methods rely on experimental validation or rule-based computational models, which are time-consuming and lagging in updating. Big language models provide a new paradigm for drug interaction prediction through semantic modeling and integrating knowledge-based reasoning on polymorphic data. In this paper, we systematically review the technical progress, application cases and technical difficulties of deep learning and big language modeling in DDI prediction. Based on the performance of five mainstream big language models, namely ChatGPT-4, DouBao, ERNIE Bot, Kimi Chat, and DeepSeek-R1, in the verifiable drug combination DDI analysis task, we explore the value of the application of big language models in drug interaction prediction and provide an outlook on the future development trend, and system development, individualized prediction optimization, and standardized system development, individualized prediction optimization, and the construction of a standardized DDI prediction framework, with a view to promoting the translation process of LLM from the laboratory to the clinic.