Enhanced AlphaFold capabilities enable large protein structure analysis
Researchers at Linköping University have developed an improved version of the AI protein structure prediction tool AlphaFold, enabling analysis of larger protein complexes and integration of experimental data. The advancement marks a significant step forward in protein research and drug development.
Proteins, the fundamental building blocks of biological processes, exist in countless variations within living organisms, controlling everything from muscle function to oxygen transport. These complex molecules comprise chains of amino acids that can range from 50 to several thousand units in length, forming intricate three-dimensional structures that determine their function.
Development of AF_unmasked
The team at Linköping University has created AF_unmasked, an enhanced version of AlphaFold that addresses previous limitations of the original software. The new tool demonstrates particular prowess in handling very large protein compounds and can effectively process experimental and incomplete data sets.
“We’re giving a new type of input to AlphaFold. The idea is to get the whole picture, both from experiments and neural networks, making it possible to build larger structures. But you can also have a draft of a structure that you feed into AlphaFold and get a relatively accurate result,” explains Claudio Mirabello, docent at the Department of Physics, Chemistry and Biology at Linköping University.
Historical context and progression
The development of AlphaFold, released as open-source software by Deepmind in 2020, represented a major breakthrough in the field of protein structure prediction, ultimately contributing to the 2024 Nobel Prize in Chemistry. The success of AlphaFold was built upon decades of protein structure data collection, with approximately 200,000 different proteins documented since the 1970s providing essential training data for the AI system.
Notably, the Linköping team’s influence on protein structure prediction predates the current advancement. Professor Björn Wallner, who has collaborated with one of the Nobel laureates, developed a precursor to AlphaFold alongside Claudio Mirabello. Their work, particularly their innovation in encoding protein evolutionary history within neural networks, published in 2019, significantly influenced DeepMind’s development of AlphaFold.
“The possibilities for protein design are endless, only the imagination sets limits. It’s possible to develop proteins for use both inside and outside the body. You always have to find new, more difficult problems when you have solved the old ones. And within our field, finding problems is no problem,” notes Wallner.
Technical implementation
The research, published in Nature Communications on 9 October 2024 [1], demonstrates how AF_unmasked can assist researchers in refining their experiments and provide guidance for protein design. This capability has significant implications for understanding protein functions and developing new protein-based therapeutics.
The computational demands of the project were met using the Tetralith and Berzelius supercomputers at Linköping University’s National Supercomputer Centre, with funding support from SciLifeLab, the Knut and Alice Wallenberg Foundation, and the Swedish Foundation for Strategic Research.
The enhancement of AlphaFold’s capabilities through AF_unmasked represents an important advancement in the field of protein structure prediction. The ability to analyse larger protein complexes and integrate experimental data will likely accelerate the development of new therapeutic proteins and deepen our understanding of protein function in biological systems.
Reference:
1. Mirabello, C., Wallner, B., Nystedt, B., et. al. (2024). Unmasking AlphaFold to integrate experiments and predictions in multimeric complexes. Nature Communications, 15, 8724. https://doi.org/10.1038/s41467-024-52951-w
