Meta may have created a rival to AlphaFold, with the company claiming its model can make predictions 60 times faster than other state-of-the-art systems.
Researchers at Meta say they have created a massive protein-folding model that has predicted the structure of 617m proteins.
Meta said its AI team developed a database focused on proteins that are found in microbes in the soil, deep in the ocean and inside human bodies, which are among are some of the least understood on Earth.
The research team published its results in a pre-print paper. The team said they generated the predictions using a large language model, which was trained “to learn evolutionary patterns and generate accurate structure predictions end to end directly from the sequence of a protein”.
They added that this model was trained on 15bn parameters and was able to predict more than 617m metagenomic protein structures over a two-week period.
It claimed this model’s predictions are 60 times faster than other state-of-the-art systems, allowing for larger databases.
Meta has shared its database, dubbed the ESM Metagenomic Atlas, along with an API to let scientists “easily retrieve specific protein structures relevant to their work”.
“ESMFold shows how AI can give us new tools to understand the natural world, much like the microscope, which enabled us to see into the world at an infinitesimal scale and opened up a whole new understanding of life,” Meta said in a blogpost.
If Meta’s AI model is proven accurate, it could be a rival to AlphaFold, the AI developed by Google-owned DeepMind. Earlier this year, DeepMind said AlphaFold made a massive scientific breakthrough by predicting the structure of more than 200m proteins.
The company first shared details of AlphaFold at the end of 2020, when it claimed the AI system had solved the problem of protein folding. A year ago, AlphaFold was released as an open-source project and the company created the AlphaFold Protein Structure Database.
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.