Proefschrift

Lost in Chemical Space, Found in Data

Developing new medicines is one of modern science's most significant hurdles, a process marked by high costs, long timelines, and frequent failures of promising candidates.

Auteur

O.J.M. Béquignon

Datum

11 juni 2025

Links

Thesis in Leiden Repository

The immense complexity of how chemicals interact with our bodies makes it extremely challenging to predict which ones will be both safe and effective for all patients. Computational tools, particularly artificial intelligence, are now playing a crucial role in tackling these challenges, offering ways to speed up discovery and reduce failures. However, the power of these tools depends heavily on robust data and methods. This thesis delves into enhancing these computational approaches for drug discovery. The works it describes focus on creating higher-quality data resources for reliable predictions; developing smarter models to foresee drug safety (such as liver injury) and effectiveness against specific targets involved in diseases, including those altered by mutations; and innovating methods to design new drug molecules by exploring and refining their chemical structures.

• chemical space
• data curation
• de novo design
• drug-induced liver injury
• machine learning
• molecular descriptors
• predictive toxicology
• proteochemometrics
• quantitative structure activity relationship
• smiles (simplified molecular input line entry system)