- Paul Rincon
- BBC News Science Editor
Artificial intelligence was used to predict the structures of almost all proteins produced by the human body, according to a new study.
The breakthrough could drive the discovery of new drugs to treat diseases, in addition to other applications.
Proteins are essential components of living organisms. Each of our cells is packed with these “essential building blocks” of life.
Understanding the forms of proteins is critical to the advancement of medicine, but so far only the structure of a fraction of them had been elucidated.
Now researchers used a program called AlphaFold to predict the structures of 350,000 proteins of humans and other organisms.
The instructions for making human proteins are contained in our genomes, the DNA in the nuclei of human cells.
There are about 20,000 of these proteins expressed by the human genome.
Collectively, biologists refer to the complete set of these proteins as the “proteome.”
“We believe this is the most complete and accurate picture of the human proteome to date,” said Demis Hassabis, CEO and co-founder of the artificial intelligence company DeepMind, which developed AlphaFold.
“We think this work represents the Most significant contribution of artificial intelligence to the advancement of scientific knowledge to date“.
“And I think it’s a great illustration and example of the kind of benefits that artificial intelligence can bring to society.”
“We are very excited to see what the scientific community is going to do with this information,” Hassabis added.
Proteins are made up of smaller building block chains called amino acids.
These chains fold in countless different ways, forming a unique 3D shape. The shape or folding of a protein determines its function in the human body.
The 350,000 protein structures that AlphaFold predicted include not only the 20,000 contained in the human proteome, but also those of so-called model organisms frequently used in scientific research, such as E. coli bacteria, yeast, fruit flies and the rats.
This giant leap in capacity is described in a new study by researchers from DeepMind and the European Molecular Biology Laboratory (EMBL) which was published in the prestigious journal Nature.
AlphaFold was able to make a reliable prediction of the folds for 58% of the amino acids in the human proteome.
The 35.7% positions were predicted with a very high degree of confidence, twice the number confirmed by experiments.
Traditional techniques for elucidating protein structures include X-ray crystallography and cryo-electron microscopy (Cryo-EM), among others.
But neither of these methods is straightforward.
“It takes a lot of money and resources to decipher the structures,” Professor John McGeehan, a structural biologist at the University of Portsmouth in England, told the BBC.
Due to those difficulties, 3D shapes are often obtained as part of specific scientific investigations, but no project until now had systematically determined structures for all proteins produced by the body.
In fact, only the structures of 17% of the human proteome have been confirmed experimentally.
“It’s the speed, the fact that it took us six months for structure and now it takes a couple of minutes. We really couldn’t have predicted it would happen that fast, “Professor McGeehan said of AlphaFold’s predictions.
“When we sent seven sequences to predict to the DeepMind team, we had already managed to experimentally decipher the structures of two of them. So we were able to compare them with the results of Deep Mind. It was one of those moments, to be honest, where you get the creeps behind your neck. The AlphaFold structures and the ones we had confirmed experimentally were identical“.
Professor Edith Heard of EMBL said this breakthrough “will transform what we know about how life works. That’s because proteins represent the fundamental building blocks from which living organisms are made.”
“Applications are limited only by our understanding.”
Those potential applications include new treatments for diseases, crops resistant to climate change, and enzymes that can break down plastic that pollutes the environment.
Professor McGeehan’s group is already using AlphaFold data to develop faster enzymes to degrade plastic.
McGeehan noted that artificial intelligence had provided predictions for proteins whose structures could not be determined experimentally, helping speed up his project by “several years.”
Ewan Birney, director of EMBL’s European Bioinformatics Institute, told the BBC that the structures predicted by AlphaFold are “one of the most important databases on the map of the human genome.”
DeepMind partnered with EMBL to make AlphaFold protein fold predictions are openly available to the global scientific community.
Hassabis noted that DeepMind plans to expand its database to include almost every sequenced protein known to science – more than 100 million structures.
Now you can receive notifications from BBC News Mundo. Download our app and activate them so you don’t miss our best content.
Eddie is an Australian news reporter with over 9 years in the industry and has published on Forbes and tech crunch.