An AI-powered lab partner

Chemistry research often involves a lot of recurring tasks: set up reactions, check to see if they worked, make some tweaks to the conditions, rinse, and repeat.

Andrew I. Cooper and his team at the University of Liverpool are working to create a future where chemists can pass off routine tasks to robots, speeding up the process of discovery and freeing up human brain space for bigger-picture planning.

The latest iteration of their robo-chemist combines a mobile robot arm with an AI algorithm, resulting in a synthesis bot that can drive itself around the lab to carry out organic reactions, analyze the products, and use the data to make decisions much like a human chemist would—but faster (Nature 2024, DOI: 10.1038/s41586-024-08173-7).

“Once the objectives are set, the system is fully autonomous,” says Cooper in an email to C&EN.

The researchers made two bots and put them to work on three different assignments: running 2-step combinatorial small-molecule syntheses, designing supramolecular materials, and screening photocatalysts for a condensation reaction.

The robot assessed whether the reactions it carried out had succeeded using nuclear magnetic resonance spectroscopy and ultrahigh-pressure liquid chromatography/mass spectrometry. If the results passed the robots’ programmed criteria for a successful reaction, they would proceed to scale it up or try to replicate it before proceeding to the next step of the process. All the data are saved so that human researchers can review it.

“Chemistry is going full self-driving lab mode . . . and this is another example of taking the field to the next level,” says Alán Aspuru-Guzik of the University of Toronto, who also works on integrating AI and robots into the laboratory but was not involved in the work. He adds that he likes how Cooper’s team integrated multiple analytical methods into the robot’s workflow to demonstrate how machine learning models can interact with the physical world. “It really tells you how far the field has gone.”

Cooper says he and his team are continuing to expand the system’s repertoire of analytical techniques. They’re also looking at ways to automate other tedious but necessary tasks such as purification.