In a new project within the UvA MMD (Molecular and Material Design) TechHub, researchers from the Van ’t Hoff Institute for Molecular Sciences, the Swammerdam Institute for Life Sciences and Amsterdam UMC are collaborating on a new approach to drug development. The research focuses on proteins known as Rho-GTPases, which regulate pressure inside endothelial cells lining blood vessels. When the balance of these proteins is disrupted, blood vessels can begin to leak.

The scientists use computer simulations to study how candidate molecules influence the structure and activity of these proteins. Generative AI plays a key role in this process. By training AI systems on large databases containing existing drug molecules, researchers can generate entirely new molecular variations that may perform better than current compounds.

To identify which generated molecules show the most promise, the team applies a machine learning technique called Bayesian optimisation. This method helps predict which molecular changes are most likely to improve effectiveness. According to Professor of AI for Chemistry Bernd Ensing, combining simulation, generative AI and optimisation in this way is still highly experimental within the field.

A major part of the project is the direct collaboration between multiple disciplines. Chemists, AI researchers, biologists and medical specialists work together to experimentally validate the AI-generated molecules in Amsterdam UMC laboratories. The researchers also aim to accelerate development by starting from already approved drugs, potentially bypassing parts of the traditional approval process.

The project reflects how AI is increasingly becoming part of medical research, not only for analysing data but also for designing entirely new molecules and treatments. By combining machine learning with biomedical science, researchers hope to significantly reduce the time required for drug discovery and development.