Evaluating the Ability of Large Language Models in Drug-drug Interaction Network Clustering Analysis
Yue Chen, Guixia Liu, Ji Lv · Current Bioinformatics · 2026
Introduction:: Large language models (LLMs) have attracted considerable attention from both the public and academic communities due to their powerful language understanding and text generation capabilities. This has led to the emergence of diverse applications in biomedicine. Methods:: In this study, we investigated the potential of general LLMs (i.e., ChatGPT5.2, DeepSeek3.2, Gemini3.0) in drug-drug interaction (DDI) network clustering analysis. Specifically, we evaluated whether LLMs can produce high-quality clustering results when guided by structured prompts incorporating drug information and DDI data. Results:: Our results demonstrate that general LLMs can perform effective DDI network clustering analysis. When guided by drug information, LLMs can reconstruct known antibiotic classes. When clustering was performed using only DDI interaction information, the models achieved high monochromaticity. Discussion:: The findings suggest that LLMs can leverage domain knowledge embedded in their training corpora to support biomedical data analysis tasks. However, issues related to stability and reproducibility should be considered. Conclusion:: LLMs offer a cost-effective and accessible approach for solving domain-specific biomedical challenges.