Happy Nice Mall

In December 2020, DeepMind surprised the world of biology. Solves a 50-year-old challenge with AlphaFold, an AI tool that predicts the structure of proteins. The London-based company last week all the details of that vehicle and published the source code.

Now the firm has announced that it has used its artificial intelligence to predict the shape of almost every protein in the human bodyas well as forms of hundreds of thousands of other proteins found in 20 of the most studied organisms, including yeast, fruit flies, and mice. This breakthrough could allow biologists around the world to better understand diseases and develop new drugs.

So far, the treasure consists of a newly estimated 350,000 protein structures. DeepMind says it will predict and release more than 100 million structures over the next few months – more or less all proteins known to science.

“Protein folding is an issue I’ve been dealing with for over 20 years,” says Demis Hassabis, co-founder of DeepMind. “It’s been a huge project for us. I would say it’s the biggest thing we’ve ever done. And in a way it’s the most exciting because it has to have the biggest impact in the world outside of AI.”

Proteins are made of long strands of amino acids that twist themselves into intricate knots. Knowing the shape of a protein’s knot can reveal what that protein does; this is crucial for understanding how diseases work and for developing new drugs or identifying organisms that can help fight pollution and climate change.

The database should make life easier for biologists. AlphaFold may be available to researchers, but not everyone will want to run the software themselves. “It’s much easier than taking a structure from a database and running it on your own computer,” says David Baker of the University of Washington Institute for Protein Design, who runs a lab that makes his own tool. to predict the structure of a protein called roseTTAFcoat and is based on AlphaFold’s approach.

For the past few months, Baker’s team has been working with biologists trying to figure out the shape of the proteins they’ve been studying before. “There’s a lot of really cool biological research that’s been accelerated,” he says. A public database containing hundreds of thousands of ready-made protein forms should be an even greater accelerator.

“It looks surprisingly impressive,” says synthetic biologist Tom Ellis, who studies the yeast genome at Imperial College London, who is excited to try out the database. But he points out that many of the predicted shapes have yet to be confirmed in the lab.

In the new version of AlphaFold, the predictions come with a confidence score that the tool uses to mark how close it thinks each predicted shape is to reality. Using this measurement, DeepMind found that AlphaFold predicted shapes for 36% of human proteins with an accuracy down to the level of individual atoms. Hassabis says this is good enough for drug development.

Previously, only 17% of the proteins in the human body could be identified in the laboratory after decades of work. If AlphaFold’s predictions are as accurate as DeepMind says, the tool has doubled that number in just a few weeks.

Even inaccurate predictions at the atomic level are still useful. AlphaFold predicted a shape for more than half of the proteins in the human body, which should be good enough for researchers to understand the protein’s function. The rest of AlphaFold’s current estimates are either wrong or for the third of proteins that have no structure until they bind to other proteins in the human body. “Floppy disks,” Hassabis says.

“It’s impressive that it can be implemented at this level of quality,” says Mohammed AlQuraish, a systems biologist at Columbia University who has developed his own software to predict protein structure. He also points out that having structures for most of the proteins in an organism will make it possible to study how these proteins work as a system, not just in isolation. “I think that’s the most exciting thing,” he says.

DeepMind publishes its tools and forecasts for free and will not say if it has plans to monetize them in the future. However, this does not exclude the possibility. To set up and run the database, DeepMind is partnering with the European Molecular Biology Laboratory, an international research institution that currently houses a large database of protein information.

For now, AlQuraishi can’t wait to see what the researchers do with the new data. “It’s pretty amazing,” he says, “we didn’t think any of us would be here so soon. It doesn’t make sense.”