Blog

In the past few months, Baker’s team has been working with biologists who were previously stuck trying to figure out the shape of the proteins they were studying. “There are a lot of interesting biological studies that are really accelerating,” he says. A publicly available database containing hundreds of thousands of ready-to-use protein forms should be an even bigger accelerator.

“It looks amazingly impressive,” says Tom Ellis, a synthetic biologist at Imperial College London who studies the yeast genome, who is delighted to try the database. But he cautions that most of the predicted shapes have not yet been tested in the laboratory.

Atomic precision

In the new version of AlphaFold, predictions have a confidence score, which the tool uses to indicate how close it thinks each predicted shape is to the actual shape. Using this measure, DeepMind found that AlphaFold predicts the shapes for 36% of human proteins down to the level of individual atoms. That’s enough for drug development, Hassabis said.

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

Even predictions that are not completely accurate at the atomic level are still useful. AlphaFold predicted the shape of more than half of the proteins in the human body, which should be good enough for researchers to determine the function of the protein. The rest of AlphaFold’s current predictions are either wrong or refer to a third of the proteins in the human body that have no structure at all until they bind to others. “They are flexible,” says Hassabis.

“The fact that it can be applied at this level of quality is impressive,” says Mohammed Al Quraish, 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 proteins in the body will allow us to study how these proteins work as a system, and not just individually. “This is what I find most exciting,” he says.

DeepMind releases its tools and forecasts for free and does not say if it plans to make money from them in the future. However, this does not rule out the possibility. To build and run the database, DeepMind has partnered with the European Laboratory for Molecular Biology, an international research institute that already has 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 impressive,” he says. “I don’t think any of us thought we’d be here that quickly. It’s overwhelming. “