- Home
- News
- General News
- Protein decoding...
Protein decoding recognised in 2024 Nobel Prize in Chemistry
25-10-2024
This year, the Nobel Prize in Chemistry was awarded to David Baker, John Jumper, and Demis Hassabis for their pioneering work on ‘computational protein design’ and ‘protein structure prediction’, highlighting the advancements Artificial Intelligence brings to this field. The ESRF has been actively involved in breakthrough research on these subjects, collaborating notably since 2022 with teams that include Baker and Jumper.
Share
“We are living a revolution in protein research, with Artificial Intelligence playing a key role in deciphering up-to-now unsolvable structures and designing entirely new proteins from scratch, a process known as de novo protein design”, explains Montserrat Soler López, head of the structural biology group at the ESRF.
Designing proteins with specific properties
David Baker, a co-recipient of this year’s Nobel Prize in Chemistry, focuses his research on protein design. His collaboration with the ESRF dates back several years, reaching a significant milestone in 2022 when he used the ESRF’s advanced resources to design protein-binding proteins based solely on their target structure, as published by Cao et al. in Nature.
More recently, in a 2024 Nature paper by Goverde et al., Baker and colleagues applied deep learning to design new proteins and used the ESRF to validate experimentally their predictions. This study focused on G-protein-coupled receptors (GPCRs), a class of membrane proteins essential to cell function but challenging to study due to their complex structure and limited experimental accessibility.
The team engineered new versions of these membrane proteins that, unlike the 'natural’" counterparts, are soluble in water. This shift in solubility was achieved by modifying the protein’s hydrophobicity, which determines the function of the protein.
To confirm that their proteins folded as predicted, the team solved their 3D structures using X-ray crystallography at the ESRF and other facilities. The results showed that the computational models were exceptionallyaccurate, capturing not only the overall fold but also the fine details of sidechain conformations, which are essential for designing functional proteins.
“Structural characterisation serves as a validation tool for these models and is, therefore, an integral part of the development of computational models of proteins, bridging the gap between theoretical predictions and real-world molecular behaviour”, explains Soler López.
The ability to create membrane-soluble protein analogues with specific functions could significantly impact drug discovery strategies.
Predicting structures
The ESRF’s contributions extend well beyond protein design. Luca Jovine from the Karolinska Institutet, a long-term ESRF user, has worked closely with John Jumper, co-recipient of this year’s Nobel Prize in Chemistry, developer of Alphafold and CEO of Google Deepmind, to reveal the structure of proteins that are particularly challenging to study.
Jumper has co-authored several scientific publications with Jovine and ESRF beamline scientist Daniele de Sanctis. In one of these studies, the team focused on determining the structure of glycoprotein 2 (GP2), a molecule produced in the pancreas and the intestine that plays a critical role in counteracting bacterial infection at the level of the gastrointestinal system, similar to the function of uromodulin (UMOD).
Jovine and de Sanctis faced challenges to obtain sufficiently accurate experimental phases to resolve the structure of GP2. However, Jumper and his team at Deepmind successfully predicted the structure using the data provided by Jovine. “The structure of what we identified as the decoy module of UMOD and GP2 was expected to be unlike any known protein”, says Jovine. “It turned out that, although AlphaFold had never"'seen’ such a fold before, its predictions were remarkably accurate. This enabled us to effectively interpret the GP2 X-ray crystallographic data and fit low resolution cryo-EM maps of UMOD — both by itself and in complex with FimH, the protein that bacteria use to bind high-mannose sugar chains. Whilst AlphaFold does not currently predict sugar structures, combining its predictions with experimental data allowed us to achieve the best of both worlds”, explains Jovine.
“We were amazed by AlphaFold’s performance, which has been essential in achieving these results. There is no doubt that machine learning is already fueling the next leap in structural biology and positioning itself as an invaluable complement to X-ray crystallography, cryo-EM and bio-SAXS studies”, concludes Daniele de Sanctis.
References:
Protein design
Goverde, C.A., et al. Nature 631, 449–458 (2024). https://doi.org/10.1038/s41586-024-07601-y
Chidyausiku, T.M., et al, Nat Commun 13, 5661 (2022). https://doi.org/10.1038/s41467-022-33004-6
Cao, L., et al. Nature 605, 551–560 (2022). https://doi.org/10.1038/s41586-022-04654-9
Structure prediction
Moi, D., et al. Nat Commun 13, 3880 (2022). https://doi.org/10.1038/s41467-022-31564-1
Text by Montserrat Capellas Espuny