A new method for the targeted design of molecular structures using AI.
The engineering of novel molecules and materials can lead to significant improvements in industrial processes, drug discovery, and optoelectronics. The search for new molecules and materials can be likened to searching for needles in a haystack. Chemical space contains tens of thousands of molecules.
This is more molecules than the number of stars in the known universe. Researchers at Leipzig University and University of Warwick have developed a new method to allow the targeted design of specific molecules and materials. Their research results have been published in Nature Computational Science.
Researchers used a variety of artificial intelligence methods to combine their experiments. Julia Westermayr, assistant professor at Leipzig University’s Wilhelm Ostwald Institute of Physical and Theoretical Chemistry, is the first author.
She explains that one model was able to predict the quantum chemical properties of molecules while the other model could learn how these molecules are made. This allows for high accuracy in screening properties. Conventional methods of calculating quantum mechanical property are slow and expensive and use a lot of computational power.
The researchers used both models to create new molecules and filter them according certain properties. This iterative process involves repeated steps until a target is achieved or certain criteria are met. “In each round the design model learned how best-suited molecules were constructed and so specifically predicted molecules with optimal properties in the next round,” Westermayr says.
In 2022, the profits of oil giants in Europe and the USA doubled
Rhyan Barrett, an intern at the University of Warwick in England funded by Artificial Intelligence and Augmented Intelence for Automated Investigations for Scientific Discovery (AI4SD), laid the foundation for this study.
New Method for Targeted Design of Molecules With AI – Technology Networks https://t.co/2pRCmCvgk7
— Emma Parkes (@emmafparkes) February 9, 2023
Optimized properties for molecular structures
Rhyan Barrett says, “We were especially surprised that artificial intelligence was able to find patterns within the data that led us to optimized properties.” The researchers were able to optimize multiple properties. The method can be used to find Pareto optimal solutions. Pareto-optimal solutions are those that combine several properties to solve a problem. Each property can only improve if the other one gets damaged.
This method was used to create functional organic molecules that are relevant to optoelectronics. These new, more efficient materials can be used in areas like the solar energy industry and LED lighting. They also have data storage capabilities. This new technology can be used in other areas.
The development of active ingredients that have specific, targeted properties and are effective against certain diseases is another area of potential application. Environmental engineering can also use molecular design to create new methods for purifying water and air. Biotechnology is based on the creation of specific molecules that perform specific functions.
