[ad_1]
A new artificial intelligence program easily predicts the structure of protein complexes, such as the interleukin-12 (blue) immune signal bound to its receptor.
Ian Haydon / Institute for Protein Design
By Elizabeth Pennisi
Proteins are the servants of life, working alone or together to build, manage, fuel, protect and eventually destroy cells. In order to function, these long chains of amino acids twist, bend, and intertwine into complex shapes that can be slow, if not impossible, to decipher. Scientists dreamed of simply predicting the shape of a protein from its amino acid sequence, an ability that would open up a world of information about how life works. “This problem has been around for 50 years; a lot of people have broken their heads, ”says John Moult, structural biologist at the University of Maryland, Shady Grove. But a practical solution is within their grasp.
Several months ago, in a result hailed as a turning point, computational biologists showed that artificial intelligence (AI) can accurately predict the shapes of proteins. This group describes their online approach in Nature today. Meanwhile, David Baker and Minkyung Baek from the University of Washington, Seattle, and their colleagues present their approach to online AI-based structure prediction in Science. Their method works not only on simple proteins, but also on protein complexes.
Baker and Baek’s method and computer code have been available for weeks, and the team has already used it to model more than 4,500 protein sequences submitted by other researchers. Savvas Savvides, a structural biologist at Ghent University, had tried six times to model a problematic protein. He says Baker and Baek’s program, called RoseTTAFold, “paved the way for a structural solution.”
In the fall of 2020, DeepMind, a UK-based AI company owned by Google, wowed the field with its structure predictions in a biennial competition. Called Critical Assessment of Protein Structure Prediction (CASP), the competition uses newly determined structures using laborious laboratory techniques such as X-ray crystallography as benchmarks. DeepMind’s program, AlphaFold2, has done “some really amazing things [predicting] protein structures with atomic precision, ”explains Moult, who organizes the CASP.
But for many structural biologists, AlphaFold2 was a tease: “Incredibly exciting but also very frustrating,” says David Agard, structural biophysicist at the University of California, San Francisco. In mid-June, 3 days after Baker Lab released their RoseTTAFold preprint, Demis Hassabis, CEO of DeepMind, tweeted that details of AlphaFold2 were being reviewed in a post and the company would provide. “broad free access to AlphaFold for the scientific community”. Nature has now hastened to publish this article to coincide with the Science paper. “It’s appropriate that he doesn’t come out after ours because our work is really based on their advancements,” Baker said.
DeepMind’s 30-minute presentation at CASP was enough to inspire Baek to develop his own approach. Like AlphaFold2, it uses AI’s ability to discern patterns in large databases of examples, generating increasingly informed and precise iterations as it learns. When given a new protein to model, RoseTTAFold proceeds along several “tracks”. The amino acid sequence of the protein is compared with all similar sequences in the protein databases. Another predicts the pairwise interactions between amino acids within the protein, and a third compiles the putative 3D structure. The program bounces around the tracks to refine the model, using the output of each to update the others. DeepMind’s approach has only two tracks.
Gira Bhabha, a cell and structural biologist at New York University School of Medicine, says both methods work well. “The advances from DeepMind and Baker Laboratories are phenomenal and will change the way we can use protein structure predictions to advance biology,” she says. A spokesperson for DeepMind wrote in an email: “It’s great to see examples like this where the protein folding community is relying on AlphaFold to work towards our common goal of increasing our understanding. of structural biology.
But AlphaFold2 solved the structures of only proteins, while RoseTTAFold also predicted complexes, such as the structure of the immune molecule interleukin-12 locked on its receptor. Many biological functions depend on protein-protein interactions, explains Torsten Schwede, structural computational biologist at the University of Basel. “The ability to manage protein-protein complexes directly from sequence information makes it extremely attractive for many questions in biomedical research. “
Baker admits that the structures of AlphaFold2 are more precise. But Savvides says the Baker Lab’s approach better captures “the essence and peculiarities of protein structure,” such as identifying chains of atoms protruding from the sides of the protein – key features of interactions between proteins. protein. Last year, AlphaFold2 needed a lot of computing power to run, more than RoseTTAFold. “Now it looks like they’ve sped up their method since CASP14, and it’s now comparable to RoseTTAFold,” Baek says.
From June 1, Baker and Baek began to question their method by asking researchers to send in their most baffling protein sequences. Fifty-six scrapers arrived in the first month, and all have now predicted structures. The Agard group sent an amino acid sequence with no known similar proteins. Within hours, his group recovered a protein model “that probably saved us a year of work,” Agard says. Now he and his team know where to mutate the protein to test ideas on how it works.
Because Baek and Baker’s group has posted their computer code to the web, others can improve it; the code has been downloaded 250 times since July 1. “Many researchers will build their own methods of structure prediction on Baker’s work,” says Jinbo Xu, a computational structural biologist at the Toyota Technological Institute in Chicago. Hassabis claims that its computer code is now also open source. Thanks to the work of both groups, progress should now be rapid, says Moult: “When there is a breakthrough like this, 2 years later, everyone is doing it as well if not better than before. “
[ad_2]
Source link