Thanks to artificial intelligence: researchers find new method for targeted design of molecules

Scientists from the University of Leipzig and the University of Warwick have now succeeded in developing a new method that enables the targeted design of molecules and materials with specific properties. They have now published their research results in the journal „Nature Computational Science“.

The researchers combined various artificial intelligence methods in their experiments. First author Junior Professor Dr Julia Westermayr from the Wilhelm Ostwald Institute for Physical and Theoretical Chemistry at Leipzig University explains: "One model has learnt to predict the quantum chemical properties of molecules, while the other has learnt how these molecules are structured.The first model is necessary to enable the screening of properties with high accuracy, as conventional methods for calculating quantum chemical properties are very time-consuming and computationally intensive. In an iterative process, in which steps are repeated until a certain goal is achieved or certain criteria are met, the researchers then used both models to generate new molecules and filter them according to certain properties.

In each round, the design model learnt how the most suitable molecules are structured and thus specifically predicted molecules with optimised properties in the next round,

says Westermayr.

The basis for the study was laid by Rhyan Barrett during an internship at the University of Warwick in England, which was funded by the network „Artificial Intelligence and Augmented Intelligence for Automated Investigations for Scientific Discovery“ (AI4SD). It brings together researchers working on the use of cutting-edge artificial and augmented intelligence technologies to push the boundaries of scientific discovery.

Molecules with optimised properties

„We were particularly amazed that we were able to use artificial intelligence to find patterns in the data that led to optimised properties,
says Rhyan Barrett. Finally, the researchers succeeded in optimising multiple properties. This makes it possible to use the method to find Pareto-optimal solutions. A Pareto-optimal solution exists when the solution of several optimised properties has been found and the individual properties can only become better if another property becomes worse.

The developed method was used to design functional organic molecules that are relevant for optoelectronics. These novel materials with higher efficiency could be used in the solar energy sector, for LED lighting, display technology, data storage, sensor technology and optical fibres in communication technology, among others. The new method can also be applied to other areas. Possible other areas of application include the development of active ingredients for new drugs with specific, improved properties that are effective against certain diseases. Molecular design can also be used in environmental sciences to develop new processes for cleaning waste water and air. In biotechnology, the development of new biocatalysts and enzymes is based on the design of molecules with specific functions.