ESMFold2 predice 1.1 mil millones de estructuras de proteínas

ESMFold2, a new AI tool, has generated an atlas of over 1.1 billion predicted protein structures, vastly expanding the known protein universe. The atlas eclipses the AlphaFold Database by more than 800 million entries. ESMFold2 is based on a 'protein language' model that was trained on billions of proteins from across the tree of life, including 'metagenomic' sequences from soil, ocean, and other environments.

Performance Comparison ESMFold2 outperforms existing methods, including AlphaFold3, at determining the correct structure of complexes of interacting proteins – including antibody molecules binding to their antigen molecular targets. In the preprint, the researchers describe how they used ESMFold2 to design new antibodies and other proteins that can strongly attach to proteins implicated in cancers and immunological conditions.

Atlas and Database The newly released atlas contains 1.1 billion predicted protein structures as well as information on the sequences of 6.8 billion proteins. Most of these come from metagenomic sequences that had been only poorly characterized. The atlas should be 'an extraordinary resource for biology,' says Gemma Atkinson, a computational biologist at Lund University in Sweden. ESMFold2's predictions of interacting proteins are impressive, but not all that surprising, given earlier gains by proprietary models like Isomorphic Labs' model. The fully open-source nature of ESMFold2, with no restrictions on commercial use, means that it could find wide use. As the field of protein-folding AI continues to evolve, ESMFold2's ability to predict the structure of proteins that are very different from those already known will be crucial to its long-term impact.

Source: New protein-folding AI vastly expands on Alphafold's efforts
Domain: scientificamerican.com