Improving Drug-Drug Interaction Prediction via In-Context Learning and Judging with Large Language Models

He Qi^1,2Xiaoqiang Li³Chengcheng Zhang^4*Tianyi Zhao^2,5*

• ¹Center for Drug Evaluation and Inspection for Heilongjiang Province, Harbin, Jilin Province, China
• ²School of Medicine and Health, Harbin Institute of Technology, Harbin, Jilin Province, China
• ³Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences (CAS), Suzhou, Jiangsu Province, China
• ⁴Faculty of Computing, Harbin Institute of Technology, Harbin, Heilongjiang Province, China
• ⁵Harbin Institute of Technology Zhengzhou Research Institute, Zhengzhou, China

Large Language Models (LLMs), recognized for their advanced capabilities in natural language processing, have been successfully employed across various domains. However, their effectiveness in addressing challenges related to drug discovery has yet to be fully elucidated. In this paper, we propose a novel LLM-based method for drug-drug interaction (DDI) prediction, named DDI-JUDGE, achieved through the integration of judging and ICL prompts. The proposed method outperforms existing LLM approaches, demonstrating the potential of LLMs for predicting DDIs. We introduce a novel in-context learning (ICL) prompt paradigm that selects high-similarity samples as positive and negative prompts, enabling the model to effectively learn and generalize knowledge. Additionally, we present an ICL-based prompt template that structures inputs, prediction tasks, relevant factors, and examples, leveraging the pre-trained knowledge and contextual understanding of LLMs to enhance DDI prediction capabilities. To further refine predictions, we employ GPT-4 as a discriminator to assess the relevance of predictions generated by multiple LLMs. These individual results are then combined using a weighted fusion method to improve overall prediction accuracy. We compared with five other state-of-the-art LLMs respectively in zero-shot and few-shot scenarios, and achieved the best performance in both cases. Our approach demonstrates promising results and highlights the potential of LLMs in advancing drug discovery. The code for DDI-JUDGE is available at https://github.com/zcc1203/ddi-judge.