In collaboration between Ludwig Maximilian University, the Swiss Federal Institute of Technology in Zurich, and Roche Pharma Research and Early Development in Basel, a research team used artificial intelligence to create an innovative method to predict the optimal synthesis and composition of drug molecules.

The new approach is taken, as detailed in paper A research study published in the journal “Nature Chemistry” aims to significantly reduce the number of clinical trials needed, thus enhancing the efficiency and sustainability of the pharmaceutical industry.

The challenge in medicinal chemistry is to modify the structures of active pharmaceutical ingredients composed mainly of carbon and hydrogen atoms, to which functional groups are attached for specific biological functions. Changing and adding these functional groups is necessary to achieve new or improved medical effects. One way to activate the structure is a borolation reaction, in which a chemical group containing boron is attached to a carbon atom in the arrangement. Controlling this process in the laboratory has proven difficult.

The research team developed an artificial intelligence model that was trained on data from reliable scientific sources and experiments conducted in an automated laboratory at Roche Pharma, and the accuracy of predictions improved when considering the three-dimensional information of the raw materials, which goes beyond just their two-dimensional chemical formulas as in classical chemistry.

The team applied the AI-based method to identify locations in existing active ingredients where additional active combinations could be included, which facilitates the rapid development of new and more effective versions of known active ingredients.

The study indicates the possibility and importance of using artificial intelligence for the purpose of developing various drug formulations in optimal ways and discovering new medicines with higher effectiveness.