Researchers at Shanghai Jiao Tong University have developed AutoDNA, a multi-agent AI system capable of designing and executing complex DNA experiments based on a single natural-language prompt. The study, led by Haibo Chen, Mingyu Wu, Zhaoguo Wang, and Jiabin Wang, was published in Nature on November 4, 2026.
The system operates by coordinating specialized AI agents that manage the planning and execution of laboratory tasks. Physical operations are handled by robotic arms that interact with more than 20 types of laboratory equipment, performing tasks such as DNA synthesis and sequencing.
In performance tests, AutoDNA achieved expert-level yields in enzymatic DNA synthesis. During a data storage experiment, the system successfully encoded and retrieved digital information, completing 11,382 hardware steps over approximately 203 hours.
The platform also demonstrated clinical utility by correctly identifying a KRAS G13D mutation in a real-world sample. While the system automates end-to-end workflows, human involvement remains necessary for equipment maintenance and restocking chemical reagents.
A preprint of the research was initially released on arXiv in July 2025. The authors note that while the system shows potential for genetic testing and data storage, it is not currently intended for clinical hospital use.
