AI, Robots Revolutionize Chemical Experiments

A research team from the Ulsan National Institute of Science and Technology (UNIST) has developed a platform that can simultaneously experiment with thousands of chemical reaction conditions using artificial intelligence (AI) and robots to create new chemical compounds. It is expected to aid in the development of new drugs and the discovery of new materials.

On Sept. 25, a team led by Prof. Bartosz Grzybowski from the UNIST Department of Chemistry, who is also director of the Center for Artificial Intelligence and Robotics-aided Synthesis at the Institute for Basic Science, announced the development of an AI and robot platform that can quickly and precisely experiment with and generate chemical compounds.

This platform combines artificial intelligence and robotics to simultaneously experiment with thousands of chemical reaction conditions, while precisely mapping the results and selectively generating desired substances. The research results were published online in the international journal Nature at 12:00 AM (KST) on Sept. 25.

Chemical reactions are often explained in simple one-line equations like “A and B react to produce C.” However, in reality, chemical reactions can yield completely different results even when using the same materials if the quantity or temperature is slightly changed, and the process takes the form of a complex network with multiple pathways.

In response, Prof. Bartosz Grzybowski’s team self-produced an AI and robot platform capable of automatically conducting about 1,000 chemical experiments, analyzing the experimental data to draw a precise map of the complex network-shaped chemical reaction process. They visualized how chemical reactions follow certain paths and lead to specific results, even discovering hidden chemical reaction pathways that appear under certain conditions.

Furthermore, they confirmed that even when using the same substances, changing the reaction conditions can convert them into entirely different products, and they also revealed new product substances that were previously unknown.

Notably, the research team precisely mapped and reconstructed the entire reaction network of the “Hantzsch pyridine synthesis reaction,” first reported 150 years ago and used in the manufacture of various pharmaceuticals such as antibiotics and anticancer drugs, using this platform. Through this, they additionally revealed 9 new intermediates and products that were previously unknown, in addition to the 7 known existing products.

Moreover, they synthesized metal compositions (756 types) of “Prussian blue analogues” used in secondary batteries and other applications, finding optimal combinations that show higher efficiency and precision than previously reported catalysts, and identifying 4 new products that have not been reported until now.

This research has broadened the understanding of previously unknown chemical reactions by revealing that even classical chemical reactions studied for a long time still contain unknown pathways to be solved, suggesting a new research direction to enhance the efficiency and precision of chemical synthesis.

Additionally, this research is significant in that it dramatically improved the efficiency of data accumulation essential for AI utilization. Until now, traditional chemical experiments alone did not provide sufficient data for AI learning. However, now it is expected that the exploration of unknown chemical domains and research on new substances will be accelerated by quickly accumulating new experimental and reaction data through robots and directly linking and utilizing it with AI.

Yankai JIA, the first author, said, “We will dramatically increase the speed of discovering new chemical substances by advancing the use of robots and artificial intelligence,” adding, “We plan to actually utilize the newly found molecules in new material research.”

Research Director Grzybowski stated, “Viewing chemical reactions as networks rather than straight lines will be an important turning point for future chemical research,” and added, “Through the use of AI and robots, we expect to greatly increase the efficiency and diversity of chemical synthesis and contribute to future new drug development and material innovation.”