Intelligent engineering of multicomponent drug crystals

Only one in five thousand drug molecules that are developed, reaches the market. One of the main factors for this low statistic is the fact that these drugs cannot be formulated in a way that can be easily dissolved and made readily available in the body to have the therapeutic effect required. For decades, researchers in academia and industry have been trying to improve this physical property of drugs by growing stable solid forms with other safe molecules. However, the success rate of finding a winning combination of drug and a co-former that is stable and has improved solubility is dwindling. With only eight such formulations known in the market, we need improved methods to enable higher success rates. Machine learning is now fast becoming the method that tackles such combinatorial issues. This project brings together key experimental information such as the molecule’s solubility in solvents and cutting-edge inter-molecular interaction information from synchrotron X-ray scattering techniques, together with state-of-the-art computational methods. All this information on how the drug molecule behaves in liquid state as well as in solid state will be probed using machine learning models build on known successful solid forms and will be tested to predict the formation of new solid forms.
This project would advance finding solutions to challenges in data science such as brining in different kinds of data together in a single model as well as help advance technical abilities at central facilities for the study of soft matter systems. A successful model that continues to be developed and added to, would see the production of new stable drug solid forms with improved solubility reaching the market in record times. This would also prove to be a leap towards being able to personalise medicine in a sustainable way.