The Institute for Intelligent Industrial Research (AIR) of Tsinghua University and Shuimu Molecular jointly launched and open-sourced the OpenBioMed Skills, a set of large model plugins for biomedical research. This project is the first to convert the decision-making process of biomedical experts into executable Agent Skill code, and it is simultaneously published on GitHub. The initial release includes 45 core skills covering five key areas: biochemistry and drug development, protein analysis and design, single-cell omics analysis, and data retrieval tool calls, aiming to break through the engineering barriers in the entire biomedical research and development process.
By deploying OpenBioMed Skills on the same-named open-source platform, researchers can directly call vertical domain multimodal models such as PharMolixFM, BioMedGPT-R1, and MutaPLM. Compared to general large language models like ChatGPT, this system demonstrates significant advantages in professional knowledge understanding and execution of pharmaceutical research business logic, and supports remote control of biomedical Agents via mobile apps such as Feishu, DingTalk, and WeChat.
Nie Zaiqing, Chief Researcher at Tsinghua University and Chief Scientist at Shuimu Molecular, pointed out that this action marks a shift in drug development from relying on expert experience to intelligent collaborative innovation. Under the trend of AI for Science, OpenBioMed Skills realizes direct driving of complex pharmaceutical research processes by natural language through systematic and coded workflows, providing solid infrastructure support for building the fourth generation of drug development paradigms.
