Crystallisation Screening DataFactory

A new computer platform will predict how drug molecules crystallise into specific shapes, before a single crystal is grown in a lab. This matters because the shape of a drug crystal—whether it forms long needles, flat plates, or compact blocks—directly affects how easily it can be filtered, dried, and turned into a tablet. Current methods for predicting crystal shape are often unreliable, especially when impurities are present, forcing pharmaceutical companies to run hundreds of trial-and-error experiments to find the right conditions. The project builds a "DataFactory" that combines existing physics-based models with machine learning, trained on data from automated crystallisation experiments. It will also account for how impurities alter crystal growth, a factor most prediction tools ignore. If successful, the tool could slash the time and cost of developing scalable drug manufacturing processes. Instead of weeks of lab work, a chemical engineer might run a simulation in hours to find the solvent and conditions that produce easy-to-handle crystals. This would accelerate the production of new medicines and reduce waste in pharmaceutical supply chains. The research is fundamental science—it aims to understand the mechanisms of crystal growth at the molecular level—but has a clear industrial endpoint in cleaner, faster drug manufacturing.