Meta (formerly Facebook) has taken the lead over Deep Mind, which predicted the structures of about 220 million proteins earlier this year.

Scientists at Meta (NASDAQ: META) used AI to predict the structures of approximately 600 million proteins from bacteria, viruses and other microbes that were not studied before. “Today, we announced an AI breakthrough that accelerates protein folding — the process of predicting a molecule’s shape. Our researchers have used it to create a database of the molecular structures of proteins, which are the building blocks of life on earth,” said an announcement from the company. (Nature)

The Meta team’s findings can be found in this paper. The researchers generated predictions from a large language model – an AI system that was trained on the huge volume of data to solve complicated maths problems and to generate text.

The model – ESMFold – makes its predictions 60 times faster than other current systems. The large language model was trained to “learn evolutionary patterns and generate accurate structure predictions end to end directly from the sequence of a protein.”

The model was trained on 15 billion parameters and then predicted more than 617 million protein structures over a two-week period.

“These are the structures we know the least about. These are incredibly mysterious proteins. I think they offer the potential for great insight into biology,” says Alexander Rives, the research lead for Meta AI’s protein team to Nature.

The predictions, as well as the underlying code of the model, are free to use for anyone.

“These structures provide an unprecedented view into the breadth and diversity of nature, and hold the potential for new scientific insights and to accelerate discovery of proteins for practical applications in fields such as medicine, green chemistry, environmental applications, and renewable energy,” said a Meta blog post.

Related Story: DeepMind’s AlphaFold AI Software Releases Database Of 200M Protein Structures

Image Credit: Flickr