Queen Mary University of London and AstraZeneca plc KTP 24_25 R2

A new AI platform will learn from chemical data to predict which drug molecules are most likely to succeed, cutting years off the typical development timeline. Drug discovery today is slow and expensive. A single new medicine can cost over a billion pounds and take more than a decade to reach patients, largely because most candidate molecules fail during testing. This project, a collaboration between Queen Mary University of London and AstraZeneca, builds machine learning models that analyse existing chemical and biological data to flag promising compounds early and discard unlikely ones before costly lab work begins. If the platform works, pharmaceutical companies could test fewer dead-end molecules, reduce animal testing, and bring effective treatments to market faster and at lower cost. The impact would be felt most directly in the efficiency of early-stage drug development—a part of the pipeline that patients rarely see but that determines whether a potential therapy ever reaches a pharmacy shelf. The research is applied and commercially focused, with AstraZeneca integrating the tools into its own discovery workflow. Success would not transform a single disease overnight, but it could quietly accelerate the entire pipeline for many future medicines.