Artificial intelligence (AI)-based protein structure prediction technology is undergoing revolutionary advancements. Following AlphaFold2, which won the Nobel Prize in Chemistry last year, AlphaFold-Multimer has emerged, enabling the prediction of protein complex structures with high accuracy. With the ability to analyze millions of protein complexes on a large scale, there is a growing need for tools that can analyze protein complexes quickly and efficiently.
A research team led by Professor Martin Steinegger from Seoul National University and Professor Emmanuel Levy from the University of Geneva in Switzerland announced on the 6th that they have developed "Foldseek-Multimer," which can rapidly identify the structures of protein complexes. The research results were published on the 5th in the international journal "Nature Methods."
Foldseek-Multimer checks how similar the structures of protein complexes are by overlapping them. This enables the identification of similar complex structures in millions of protein complex databases at a speed 1,000 to 10,000 times faster than conventional methods. It successfully compared billions of pairs of protein complexes in just 11 hours.
Considering that most proteins function as complexes, Foldseek-Multimer, which aids in the comparison of protein complex structures, is expected to be utilized in areas such as protein interactions, metabolic functions, and structure-based drug development.
The research team noted, "In the field of drug development, it is expected to contribute to deepening the understanding of protein interactions, accelerating protein design, and expanding new knowledge about cellular functions," adding, "It will establish itself as an essential tool for structure-based protein complex analysis." The developed Foldseek-Polymer will be provided through a user-friendly web server.
References
Nature Methods (2025), DOI: https://doi.org/10.1038/s41592-025-02593-7