(567g) Predict New Cocrystals Via Mechanochemistry

Within the pharmaceutical industry around 60% of recently developed drugs have solubility challenges while 80% of in-production line drugs are classified as being poorly soluble. Solubility is a critical property for drug performance (e.g. bioavailability) [1,2]. Cocrystal formation is one possible solution to overcome solubility problems. Pharmaceutical cocrystals, formed by combining an active pharmaceutical ingredient (API) with a cocrystal former (coformer), are often manufactured in a variety of methods that rely on the utilization of solvents. Pharmaceutical companies use between 25-100 kg solvent per kg of product, causing major water and air pollution. Hence, state-of-the-art solvent-based methods need to be replaced with more sustainable alternatives. Additionally, the formation of a cocrystal requires its components to be soluble in the same solvent. These limitations hinder the processability and development of novel pharmaceutical products.

Mechanochemistry utilizes mechanical forces to induce reactions. The application of mechanochemistry to form cocrystals without solvents is a novel, promising and sustainable method [1,3]. Scientists have exploited the features of mechanochemistry, but only in a random, non-systematic manner for the production of new cocrystals. Indeed, with the enormous number of potential API-coformer combinations, it is currently impossible to predict if two crystalline materials will cocrystallize under mechanochemical forces.

In this work, a high throughput screening approach combined with data analysis (machine learning) has been developed to predict if two crystals will cocrystallize under mechanochemical conditions. Strategically distinct pairs of APIs and crystalline coformers were screened by neat grinding (solventless) in a Retsch MM 400 vibration-mill. Powder X-ray diffraction has been applied to characterize all the samples and to determine the formation of new crystal structures. With the help of these results and molecular descriptors a machine learning algorithm (random forest model) has been successfully developed, effectively predicting the propensity of two molecules to cocrystallize under mechanochemical conditions. Additional cocrystallization events are predicted and experimentally validated. The technical feasibility of further API-coformer combinations can be estimated to create novel cocrystals without any solvent limitations, expanding the scope of mechanochemistry as a technique for drug discovery.

[1] C. B. Aakeröy and A. S. Sinha. Co-crystals. Monographs in Supramolecular Chemistry. The Royal Society of Chemistry, 2018.
[2] M. A. E. Yousef and V. R. Vangala. Pharmaceutical cocrystals: Molecules, crystals,formulations, medicines. Crystal Growth & Design, 19(12):7420–7438, 2019.
[3] S. Lou, Y. Mao, D. Xu, J. He, Q. Chen, and Z. Xu . Fast and Selective Dehydrogenative C–H/C–H Arylation Using Mechanochemistry. ACS Catalysis, 6 (6) , 3890-3894, 2016.