An international team of scientists including Kieron Burke, UCI professor of chemistry, has created a machine-learning algorithm that predicts molecular behavior. The breakthrough may aid in the development of pharmaceuticals and materials to enhance the performance of batteries, solar cells and digital displays. In a study published in Nature Communications, the researchers describe how their algorithm gathers knowledge about atomic interactions in a molecule and then uses that information to anticipate new actions. Complex atomic interactions are prescribed by quantum mechanical calculations. The research team, which also included scientists from New York University and the Technical University of Berlin, found a way to simulate chemical behavior within a molecule without having to perform quantum-level number crunching. They did so by compiling a small sample data set of the molecule in order to train the algorithm and then using it to simulate complex chemical behavior within the molecule. “Machine learning is often used for online searching, text analysis, face recognition and even suggesting movies you might like,” Burke said. “Our work applies these methods to physics and chemistry, producing algorithms that no human possibly could and allowing new science to be done, thanks to their efficiency.” The project was partially supported by grants from the U.S. Army Research Office, the National Science Foundation, the government of South Korea and the Einstein Foundation.