AI predicts the shape of nearly every protein known in science

In 2020, an artificial intelligence lab called DeepMind unveiled technology that can predict the shape of proteins – the microscopic mechanisms that drive the behavior of the human body and all other living things.

A year later, the lab shared the tool, called AlphaFold, with scientists and Unlock predictable shapes of over 350,000 proteins, including all proteins expressed in the human genome. He immediately transformed the course of biological research. If scientists can identify the shapes of proteins, they can accelerate the ability to understand diseases, devise new drugs and, otherwise, investigate the mysteries of life on Earth.

Now, DeepMind has released predictions for nearly every protein known to science. The London-based lab, which is owned by the same parent company as Google, said Thursday it has added more than 200 million predictions to an online database freely available to scientists around the world.

With this new release, the scientists behind DeepMind hope to accelerate research into the most mysterious organisms and launch a new field called metaproteomics.

“Scientists can now explore this entire database and look for patterns – associations between species and evolutionary patterns that may not have been clear until now,” said Demis Hassabis, CEO of DeepMind, in a phone interview.

Proteins start out as chains of chemical compounds, then twist and fold into three-dimensional shapes that determine how these molecules relate to others. If scientists can determine what a particular protein looks like, they can decipher how it works.

This knowledge is often a vital part of fighting disease and disease. For example, bacteria resist antibiotics by expressing certain proteins. If scientists can understand how these proteins work, they can begin to confront antibiotic resistance.

Previously, determining the shape of a protein required extensive experiments involving X-rays, microscopes and other instruments on a laboratory bench. Now, by looking at the series of chemical compounds that make up the protein, AlphaFold can predict its shape.

Technology is not perfect. But it can predict the shape of a protein with an accuracy that rivals physical experiments about 63 percent of the time, according to independent standard tests. With the prediction in hand, scientists can verify its accuracy relatively quickly.

Clement Verba, a researcher at the University of California, San Francisco, who uses technology to understand the coronavirus and prepare for similar pandemics, said the technology has “shipped up” this work, often saving months of trial time. Others used the tool as they struggled to combat gastroenteritis, malaria and Parkinson’s disease.

The technology has also accelerated research outside the human body, including efforts to improve honeybee health. DeepMind’s expanded database could help a larger community of scientists reap similar benefits.

Dr. Verba, like Dr. Hasabis, believes the database will provide new ways to understand how proteins behave across species. He also sees it as a way to educate a new generation of scholars. Not all researchers are well versed in this type of structural biology. A database of all known proteins lowers the entry barrier. “It could bring structural biology to the masses,” said Dr. Verba.