The results, published in the journal Nature, point to faster, cheaper paths to new drugs, dyes and electronic materials.
The group, based at the Institute for Basic Science in Ulsan, South Korea, built an in-house platform that pairs a horizontal robot with liquid-handling tools and software using artificial intelligence.
Instead of running one reaction at a time, the machine performs hundreds at once in tiny vials, then uses light measurements to read out what each mini-experiment produced.
The team says a complete scan that would normally take years can finish in hours.
“We showed that reactions do not proceed as a straight line from A plus B to C,” said Rafał Frydrych of the Center for Algorithmic and Robotized Synthesis at the Institute for Basic Science.
“They form a network of connections between many elements, and with our method a chemist can move through that network by changing only the starting conditions,” he told Poland's PAP news agency.
The team compares the device to electronics, specifically troubleshooting an integrated circuit. By reading the inputs and outputs, technicians can infer the wiring inside.
In the same way, they use initial conditions and end products to rebuild chemical reaction networks. Rather than a single path from ingredient to outcome, chemistry appears as a steerable web of routes.
The platform relies on fast optical checks rather than time-consuming lab analyses. Each sample is examined by its ultraviolet–visible, or UV-Vis, light spectrum, and algorithms estimate which products formed and in what amounts.
In plain terms, the system shines light through each vial, reads the chemical signature, and converts that signal into a map of the reaction’s outcomes.
The researchers call the full map a “hyperspace,” meaning the many-dimensional set of possible temperatures, concentrations, and catalysts.
Scanning well-known processes with this approach, the team uncovered hidden branches and previously missed compounds.
The project is led by Prof. Bartosz A. Grzybowski, a Polish chemist known for work at the intersection of computation and synthesis.
'This tool is meant to make chemists’ work easier'
The authors stress practical benefits. Unified reaction maps can reduce energy and raw-material use, and the consistent, repeatable data are ready for machine-learning tools.
The team has open-sourced both the hardware designs and software so labs worldwide can build the device themselves.
The setup costs about USD 25,000, according to the researchers, and can run up to a thousand reactions per day.
“This tool is meant to make chemists’ work easier,” Frydrych said. The authors argue that systematically scanning reaction “hyperspaces” can turn discovery into a more predictable, data-driven process, speeding up both the understanding of old reactions and the finding of new ones.
(rt/gs)
Source: naukawpolsce.pl