Reproduce “ChatGPT-powered Conversational Drug Editing Using Retrieval and Domain Feedback” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shengchao Liu, Jiongxiao Wang, Yijin Yang, Chengpeng Wang, Ling Liu, Hongyu Guo, Chaowei Xiao
ChatDrug, a framework developed by researchers at institutions including Mila-Québec AI Institute and Princeton University, applies large language models to interactive drug editing for small molecules, peptides, and proteins. The system demonstrated superior performance in 33 of 39 tasks by integrating domain knowledge and enabling iterative refinement, and it also provides interpretable explanations for its molecular modifications.
Key insights
- Large language models can be effectively leveraged for complex drug editing tasks by framing requests at a high level, focusing on desired property changes rather than specific structural modifications.
- Integrating external molecular knowledge and providing domain feedback via a retrieval and feedback module (ReDF) is crucial for enhancing LLM performance and ensuring chemically valid outputs.
- An iterative, conversational approach allows for progressive refinement of molecules, guiding the LLM towards optimized structures while balancing property changes with structural similarity to the original molecule.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.