The Nobel Prize in Chemistry went to an AI model (and rightly so)
Not experiments and lab coats, but computers and artificial intelligence: this year's Nobel Prize in Chemistry went to the inventors of the groundbreaking AI model, AlphaFold. This programme accurately predicts protein structures based on their genetic code—a crucial step in understanding biological functions and developing new medicines. At Leiden University, it has already become indispensable, say professors Remus Dame and Gerard van Westen.
For a long time, one of the major challenges in biochemistry was that while the genetic code reveals the building blocks of a protein, it is often difficult to determine how that protein is folded. AlphaFold has changed that. It predicts the structure of a protein based on its corresponding DNA code. ‘AlphaFold learns from all known protein structures that have been determined through experiments,’ explains Remus Dame, professor of Molecular and Cellular Chemistry. ‘As a result, it can make extremely accurate predictions.’
This is made possible by the Protein Data Bank (PDB), an extensive database in which scientists have meticulously collected the three-dimensional structures of proteins along with their corresponding genetic codes over many years. The size and consistency of this database make AlphaFold's predictions highly accurate.
Gerard van Westen in De Volkskrant
Van Westen previously shared his views on the Nobel Prize for Chemistry in the Dutch newspaper De Volkskrant: ‘Just as ChatGPT can generate entirely new texts, we use this software similarly. You can give it the command: given this protein, generate a molecule that fits into it, like a key in a lock.’
For more on this, read our In the media article (in Dutch).
‘A Nobel Prize is certainly deserved’
According to Gerard van Westen, professor of Artificial Intelligence and Medicinal Chemistry, AlphaFold has revolutionised science. ‘This recognition is certainly deserved. It’s great to see computational chemistry finally being acknowledged as a significant part of the field.’ Dame agrees: ‘The scope of this invention, both within and beyond chemistry, is enormous. AlphaFold makes complex experiments unnecessary, allowing us to take follow-up steps more quickly.’
'The scope of this invention is enormous, both within and beyond chemistry.'
AlphaFold in Leiden: faster insights and better predictions
For Dame's research group, the ability to quickly predict protein structures is incredibly valuable. ‘It helps us enormously in uncovering the function of such proteins.’ He also uses AlphaFold to predict interactions between proteins and other molecules. ‘This allows us to identify small molecules that bind to proteins and potentially block unwanted functions, generating ideas for new medicines. We also predict how different proteins bind to DNA and what they do there: for instance, transmitting signals, turning genes on or off, or folding the DNA. This provides important insights into cellular processes.’
Van Westen also makes extensive use of AlphaFold in his research group. ‘We use the predictions to simulate how proteins behave and look for new targets for drugs.’ But they also use AlphaFold to actually design new medicines: if you know what a disease-causing protein looks like, it’s much easier to design a molecule that fits it precisely. Van Westen: ‘This can even be done with limited data and offers a wealth of new possibilities for drug development.’
AI is here to stay
Both professors see a future where AI becomes increasingly integrated into the daily work of chemists. Van Westen notes, ‘Applications like AlphaFold, and models that predict the “makeability” of molecules, will become part of chemists' everyday work. They won’t replace existing methods but will offer new possibilities and valuable tools.’
AI already integrated in undergraduate education
In the undergraduate Bio-Pharmaceutical Sciences programme, several computational courses are already part of the curriculum. Van Westen: ‘Understanding what AI can and can’t do is essential knowledge for the scientist of the future.’
Dame shares this view: ‘Researchers in the future will be able to use protein structures in their work without needing in-depth knowledge. This will lead to many new insights into the structure and function of proteins. On a cellular level, AI will help us recognise patterns that provide insights into disruptions caused by diseases, drugs, or genetic differences. I expect that both fundamental research and drug development will accelerate rapidly.’
Header image: Gerard van Westen envisions the transformation of his field 'from classical chemistry to modern chemistry that integrates AI applications at the bench.' Credits: Gerard van Westen. Black-and-white image: stock photo, coloured image: AI