A team led by Prof Frank Glorius from the Institute of Organic Chemistry at the University of Münster has developed an evolutionary algorithm that identifies the structures in a molecule that are particularly relevant for a respective question and uses them to encode the properties of the molecules for various machine-learning models.
