Russian scientists have developed a digital AI-powered assistant capable of reducing the cycle of discovering new chemical reactions 180-fold — from 3.5 years to just one week. The system has already helped chemists uncover previously unknown reactions that may prove useful in the development of antitumor drugs, the press service of the Russian Science Foundation (RSF) reported.
Researchers from the N. D. Zelinsky Institute of Organic Chemistry (Moscow) have developed an algorithm that predicts the course of cycloaddition reactions — processes in which several molecules or parts of a molecule join together in a concerted fashion to form a ring, without the formation of intermediate fragments. The resulting cyclic molecules are used in antibiotics, anticancer agents, and other pharmaceuticals.
The chemists tasked the digital assistant with analyzing data on the properties and structures of 134,000 molecules. In response, the AI generated 31,000 possible reactions and, within hours, selected 205 of the most realistic candidates — those not requiring expensive or rare reagents and capable of proceeding under ordinary conditions.
The four most promising reactions were tested in the laboratory, and two successfully yielded novel, useful products. “Fragments of the synthesized molecules are found in antitumor drugs, suggesting that these new reactions could be employed in their development,” the statement noted.
“The digital assistant takes on the heavy lifting of preliminary analysis, while the human researcher makes creative decisions within a well-processed data space. Our study demonstrates that even without expensive robotic laboratories, it is possible to significantly accelerate the discovery of new reactions by leveraging accessible computational tools and expert experience,” said project lead Valentin Ananikov. According to him, the team plans to expand the methodology to other chemical processes and develop a universal AI co-expert for chemists.