Explainer: Why did protein design and structure prediction win the 2024 Nobel Prize in Chemistry? | News

What did the award winners actually do?

In the 1990s, Baker began researching how proteins fold. From these findings he developed Rosetta: computer software for predicting protein structures.

Originally, Rosetta was used to convert amino acid sequences into structures, but after the 1998 CASP competition, Baker and his team decided to use the software in reverse; A technique that eventually led them to design entirely new proteins from scratch, also known as de novo design.

To do this, they drew a protein with a completely new structure and had Rosetta figure out what type of amino acid sequence would lead to that protein. They then introduced a gene that encoded their proposed amino acid sequence into bacteria that produced the new protein.Top7. Using X-ray crystallography, they were able to determine that the protein they made had a structure very similar to the one they had originally designed.

The work of Baker and his colleagues was published in 2003 and the code for Rosetta was made available to the global research community to enable continued development of the software and new applications.

In 2010, Hassabis, a British computer scientist and AI researcher, founded DeepMind Technologies. DeepMind initially developed AI models for popular board games and, after being acquired by Google in 2014, reached a milestone in machine learning when its AlphaGo program defeated the world’s top Go player in 2016. The company then developed a computer program based on a convolutional neural network – called AlphaFold.

In 2018, AlphaFold outpaced the rest of the field in 13th place^Th CASP competition achieving 60% accuracy for its predicted protein structures. However, achieving higher accuracies presented a new challenge.

Enter Jumper, a researcher with creative ideas for improving AlphaFold. Together, Jumper and Hassabis led the work that led to AlphaFold2 in 2020, supported by Jumper’s knowledge of proteins and the innovation behind a huge breakthrough in AI – neural networks called transformers – that could find patterns in massive amounts of data more flexibly than ever before before.

When an amino acid sequence with an unknown structure is entered into the program, it searches the database for similar amino acid sequences and protein structures. The network then creates an alignment of similar sequences, sometimes from different species, and looks for correlations between them as well as possible interactions between amino acids. Using this information, AlphaFold2 can then iteratively refine a distance map – which tells you how close two amino acids are to each other in space – and perform sequence analysis. All information is then converted into a 3D structure.

AlphaFold now has more than 2 million users and has led to the prediction of 200 million protein structures.

What are the applications of this work?

Because of these breakthroughs, most monomeric protein structures can now be predicted with high accuracy, and as a result, large databases containing hundreds of millions of structures have been created. Proteins are such an important part of our biology that the ability to design them and predict their structures opens up potential applications in pharmaceuticals, nanomaterials, and the rapid development of vaccines and much more.

Does this mean the end of experimental work in this area?

There is no doubt that the development of AI tools for protein structure prediction such as AlphaFold represents an important milestone in structural biology, but they are not a replacement for experimental structure determination. Experimentally determined structures are still superior to predictions and are also needed to generate the training data sets for the next generations of AI tools and to evaluate the performance of these tools in predicting structures.

One example of the ongoing need for experimental approaches is drug design. Although determining the structure of a protein can help generate ideas about what compounds to make next, there are many other factors related to the biological activity of proteins that need to be considered, such as: B. pharmacokinetics, metabolism and toxicology that cannot currently be solved using AI.

It is much more likely that the future of structural biology lies in integrating high-throughput experimental studies with AI rather than replacing it.