DeepMind’s AI has mapped 200+ million proteins—nearly all known to scientists

DeepMind’s AlphaFold tool has mapped out the structures of nearly every protein on Earth—or, at least, the ones known to scientists.

And it’s making that data available to the scientific community for free.

The power of proteins

The news comes nearly two years after the Google-owned AI lab first introduced AlphaFold, a neural network that uses a protein’s amino-acid sequence to predict its shape.

The tool sparked a breakthrough in a 50-year-old biological mystery, since predicting a protein’s shape essentially enables researchers to predict its function. That’s big news for disease research and drug discovery, as research suggests incorrect protein folding could be the cause of degenerative disease.

What’s new: DeepMind’s update means that 200+ million protein structures are now available to scientists on the AlphaFold database, which is hosted on Google Cloud.

That’s a 200x increase from a year ago, when the database had just ~1 million protein structures—and since then, 500,000 researchers from 190 countries have used the data to do things like help break down plastic waste and advance research on neglected diseases.

Big picture: Proteins are the building blocks of all living things—humans, animals, and other organisms, including bacteria and viruses. The new data could help researchers make progress in areas like disease research, sustainability, and food insecurity, DeepMind CEO Demis Hassabis said at a press briefing. It covers, essentially, “the entire protein universe,” he said.