Nobel laureate met young researchers at the university
John Jumper, one of this year’s Nobel laureates in chemistry, visited Stockholm University on 11 December for a lecture on the work with AlphaFold.
For more photos, see the bottom of the page.
The day after the Nobel Prize ceremony and the banquette at the Stockholm city hall, Nobel laureate John Jumper visited Stockholm University to give a lecture and meet PhD students and researchers in molecular biology and biochemistry. The lecture was arranged by the PhD board at the Department of Molecular Biosciences, The Wenner-Gren Institute (MBW).
Nobel Prize to protein structure prediction
The Nobel Prize in Chemistry 2024 was awarded to, with one half to the American professor David Baker “for computational protein design”, and the other half jointly to Dr. Demis Hassabis and Dr. John Jumper, both at Google DeepMind in the UK “for protein structure prediction”.
In 2020, Demis Hassabis and John Jumper presented an AI model called AlphaFold2. Using it, they have been able to predict the structure of virtually all the 200 million proteins that researchers have identified. Among many scientific applications coming out of AlphaFold2, researchers can now better understand antibiotic resistance and create images of enzymes that can decompose plastic.
John Jumper, born in 1985, started his career as a physicist turning into theoretical chemistry before taking on his current position as a senior research scientist at Google DeepMind in London. He stated that the answer is physics, but not in the way you think – it's more complex than that. It's coevolution in combination with physics and not a single thing. That is because it is hard to say exactly what has been learned.
AlphaFold2 as a “coevolutionary process”
John Jumper described AlphaFold2 as a “co-evolutionary process” where knowledge has been incorporated from areas like biology, chemistry, computer science and physics. He also stressed that all biochemistry come out of data. There are large amounts of data – and data “can do a lot”.
He admitted that AlphaFold2 is “pretty complicated” and demonstrated how it is built by looking at sequences and using different templates. The AI model has 48 layers, but recycling information three times effectively creates a deeper network. Building in certain constraints into how information is passed inside a deep learning framework can also help the learning to reach an accurate solution faster.
New possibilities with AlphaFold3
John Jumper also talked about the next evolution known as AlphaFold3. This new AI model will be able to predict structures of all kind of molecules. He also touched upon a feature called AlphaFold Server. This is a web-service that can generate highly accurate biomolecular structure predictions containing proteins, DNA, RNA, ligands and ions, but also model chemical modifications for proteins and nucleic acids in one platform. It’s powered by the newest AlphaFold 3 model.
”Career paths are rarely linear”
After the lecture there was a coffee break when some of the PhD students and researchers took the opportunity to have a chat with John Jumper. Afterwards there was a Q&A session for the PhD students with the Nobel laureate. The questions ranged from how John Jumper ended up winning the Nobel prize to what his career path was.
“I would say that the main message we got from his answer is that career paths are rarely linear. He said we should follow our curiosity and passion rather than adhering to traditional routes. Also, to not be afraid of quitting something that does not fulfil us (he quit his first PhD project) and that sometimes the best opportunities come unexpectedly,” says Isabella Badolati, PhD student and one of the organisers of the event.
There were also questions asked about AlphaFold2 and about his work on protein folding. Some questions were technical and about the future for structure prediction.
Massive impact by AlphaFold2
Patrick Bryant introduced the lecture with John Jumper. He is an assistant professor at the Department of Molecular Biosciences, The Wenner-Gren Institute and head of a research group using AI to predict the structure of proteins from sequence information. According to him, AlphaFold2 has had a massive impact on understanding proteins' structure.
“But I think its true potential lies in protein interactions and we see this as the current focus for development from DeepMind (AlphaFold3) and others. We originally utilized AlphaFold2 for protein-protein interactions and the same MSA procedure is used in AlphaFold-multimer and AlphaFold3. In my lab, we build upon these networks to do more specialized tasks like the prediction of more difficult protein interactions and also to design new ones.”
What is your main takeaway from John Jumper's visit?
Photo: Johanna Säll
“John is a great scientist with a deep understanding of many different fields, driven by curiosity and to have a positive impact on the world – the students thought it was really great to have him here with us,” says Patrick Bryant.
Max Louski, PhD student and chair of the PhD board, adds:
“I think the main message that John Jumper gave us is to follow our curiosity despite our background, the required skills can be acquired on the way. ”
Watch the Nobel Lecture 2024 in chemistry in Aula Magna, Stockholm University
Read more on the Nobel prize in chemistry 2024
Photos from John Jumper´s visit.
Last updated: December 12, 2024
Source: Communications Office