Nobel prize in chemistry for computations of proteins

published 2024-10-10

Michal's opinion on this years Nobel prizes, as published on the UCT Prague website

The 2024 Nobel Prizes in Chemistry were awarded for two discoveries that are related, though they occurred in different contexts. David Baker from the University of Washington, USA, received the prize for the rational design and synthesis of proteins. His team's efforts spanned over twenty years, during which they worked to design new types of proteins. The second part of the Nobel Prize, awarded to Demis Hassabis and John M. Jumper from the company DeepMind, was given for the computational tool AlphaFold, which not only facilitated protein design but also fundamentally changed how scientists today think about protein structure.

Proteins are biopolymers, naturally occurring molecules composed of universal building blocks called monomers. Cells typically use 20–22 types of monomer units. Proteins are linear, non-branched chains typically made up of 300 monomers. Since the 1970s, it has been known that the sequence of monomers defines a specific three-dimensional protein structure. One of the unresolved questions was how to determine the three-dimensional structure from a known sequence of monomers. The 3D structure of a protein also determines its function, and it can only fulfill its function if it is folded correctly.

Natural proteins can be classified into several hundred classes. David Baker was interested in how to create a protein of any shape in the laboratory, outside of the known classes, and how to control certain properties, such as the electrical charge on the protein's surface. This is no easy task, as folding does not happen randomly but is based on a complex network of physical interactions between the monomers. Baker's team developed several computational methods, which came together in the Rosetta software package, long before DeepMind developed AlphaFold.

In addition to computational protein design, Baker also synthesized proteins and demonstrated that they could be used as biosensors or advanced nanomaterials.

A generation younger, Hassabis did not focus on proteins. His formal education is in computer science, and he worked as a software engineer specializing in deep neural networks. Incidentally, this year's Nobel Prize in Physics was awarded to J. Hopfield and G. Hinton for their work in this field. Hassabis worked at the London-based startup DeepMind, which was later acquired by Alphabet (which also owns the Google search engine). DeepMind became famous for its AlphaGo algorithm, which plays the board game Go using machine learning principles. It turned out that a similar strategy could be applied to the problem of predicting the three-dimensional structure of proteins.

The CASP (Critical Assessment of protein Structure Prediction) competition regularly takes place to predict 3D protein structures. In this competition, the sequences of several proteins are announced. The organizers experimentally determine the 3D structures but do not disclose them during the competition. Laboratories worldwide can then participate by predicting the 3D structures from the known sequences. After submissions close, the predictions are compared to the experimental structures, and rankings are determined. The first version of AlphaFold from DeepMind performed quite well in the 13th round of the competition in 2018. In the next round in 2020, AlphaFold2 won by a wide margin. The entire scientific community was amazed by the results and quickly realized that this was the solution to a problem that had been unresolved for about 50 years. Upon releasing their algorithm, the authors also published predictions for the three-dimensional structures of all human proteins and later expanded the database to include predictions for about 200 million known protein sequences. AlphaFold2 (and its third version) are freely available online, allowing anyone to use them. This revolution in thinking about proteins has impacted fields ranging from molecular biology to medicinal chemistry.

The third laureate is J. M. Jumper, the lead author of the paper on the AlphaFold2 algorithm. His work also has a Czech connection – Augustin Žídek, a Czech scientist, is listed among the co-authors of both the paper and the algorithm. Jumper won the Nobel Prize just a few years after completing his PhD (he finished it in 2017), and the key part of his work was published in 2021. It is not common for the Nobel committee to recognize such recent discoveries. For comparison, this year's Nobel Prize in Physiology or Medicine was awarded to scientists for the discovery of microRNA, which was first published in 1993.

The impact of the AlphaFold algorithm is massive, which is why we at University of Chemistry and Technology in Prague are working to introduce students to its functioning. As part of the core curriculum of several bachelor's programs, we have a course called Computational Chemistry, which is currently undergoing revisions. The goal is to give students hands-on experience with machine learning algorithms like AlphaFold. The Nobel committee has aligned with our goals by awarding this year's chemistry prizes, and I believe it will make motivating students much easier.