Shanghai Jiao Tong University Team Builds AI Science Lab AutoDNA, Featured in Nature

Many existing AI science labs rely on experts to set up instruments and rules in advance. AI only handles small parts like paper search or data analysis, so they struggle with complex biology experiments.

Teams from Shanghai Jiao Tong University, led by Chen Haibo, Wang Zhaoguo, Fan Chunhai and Wang Fei, built an AI-native science lab called AutoDNA. It can run molecular biology experiments on its own, and the work was published in Nature and featured on its homepage.

The system co-designs models, experiments and instruments. Multiple AI agents break tasks into steps like paper search, plan design, code generation and result analysis. Each instrument is turned into an atomic service, so AI can call devices across the lab.

The hardware supports micron-level positioning and microliter liquid handling, running on a control terminal based on OpenHarmony. It has been tested in nucleic acid detection, DNA synthesis and cancer detection.