Science magazine has revealed its breakthrough of the year: Artificial intelligence discoveries that predict how proteins fold, forged by researchers at the University of Washington’s Institute for Protein Design and Alphabet’s DeepMind.
The international academic journal annually bestows the award on the year’s biggest finding across fields, from physics to astronomy to biology. The UW researchers behind this year’s award developed a tool called RoseTTAfold that stunned scientists across the physical and life sciences with its speed and accuracy at predicting how proteins form three-dimensional shapes.
“The breakthrough in protein folding is one of the greatest ever in terms of both the scientific achievement and the enabling of future research,” wrote H. Holden Thorp, a biochemist and the Editor-in-Chief of Science in an editorial.
Researchers traditionally relied on time-consuming laboratory methods to assess the shapes of proteins, which power much of the body’s functions, from metabolism to cell division.
For decades scientists sought a computational solution to this problem. But the task was daunting. With 20 amino acid building blocks that fit together like beads on a string, the options for how an individual protein might fold are numerous. Folding depends on multiple molecular interactions within the protein and its environment, which are constantly shifting during the folding process.
The IPD revealed its solution to this problem in a publication in Science in July. The researchers’ deep learning tool, trained on known protein structures, predicted the folding of hundreds of previously-uncharted proteins. These included proteins linked to cancer cell growth, inflammation disorders, and other conditions. DeepMind published its approach in Nature the same week, and released its method to the scientific community.
“Now structures can be obtained for samples that defy experimental methods, and moreover, in labs that can’t afford the experimental approaches. It is truly protein structure for all,” said Thorp.
Computational chemist Minkyung Baek, first author on the IPD study, led the development of RoseTTAFold in collaboration with colleagues in the lab of IPD head David Baker and researchers at institutions in Victoria, B.C., South Africa and the United Kingdom.
“All areas of computational and molecular biology will be transformed,” Baker told Science in a feature story describing the vast implications of the work for understanding biology and accelerating drug development. Science magazine, known more for its muted assessments of new research, said the breakthrough, “offers a view of the dance of life as never seen before, a panorama that will forever change biology and medicine.”
RoseTTAFold and DeepMind’s AlphaFold are already being leveraged by industry for the development of new therapeutics. Startups in the field include IPD spinout Cyrus Biotechnology, which can engineer therapeutic proteins with properties such as greater stability in the body. The Seattle-based company is developing a protein-based therapeutic that’s built to bind to disarm COVID-19, and it has 90 industry partnerships for other projects.
Other IPD spinouts include Seattle-based companies Neoleukin Therapeutics, which designs anti-cancer therapeutic proteins from scratch, A-Alpha Bio, and Icosavax, which went public this year and is advancing an engineered COVID-19 vaccine. In November, Alphabet launched drug discovery startup Isomorphic Labs to build off of DeepMind’s protein folding research.
“Knowing a protein’s structure is the first step to understanding its function, drug discovery, and many other applications,” Baek told GeekWire.
Scientists in the Baker lab continued their momentum with new findings through the fall. In November, IPD researchers and their colleagues uncovered the structures of a large set of protein complexes, consisting of multiple proteins that act together in the body. And the researchers are designing new proteins not seen in nature, a step towards building bespoke proteins with specific properties.
Science also identified runner-ups for breakthrough of the year:
- Techniques to analyze DNA in ancient soils, which is yielding new insights into human and animal evolution.
- Insights into the composition of the planetary core of Mars by NASA’s InSight lander.
- Advances in fusion energy edging closer to the goal of yielding more energy in a reaction than the amount needed to spark it. Science also noted the growth in private fusion projects, such as British Columbia’s General Fusion.
- The use of CRISPR gene editing to fix genes inside the body in two hereditary diseases.
- Advances in techniques to understand early embryonic development, including a method to grow mouse embryos outside the mother for 11 days.
- The development of COVID-19 antivirals, such as Merck’s molnupiravir and a pill by Pfizer currently under review by the U.S. Food and Drug Administration. Fred Hutchinson Cancer Research Center was one of the molnupiravir clinical trial sites.
- The assessment of psychedelics to treat psychiatric conditions, which have shown signs of promise in early clinical trials of MDMA and psilocybin. University of Washington physician Anthony Back similarly recently initiated a trial of psilocybin in health care workers.
- Advances in particle physics finding that a particle called a muon is slightly more magnetic than scientists had predicted — suggesting the existence of new, unknown subatomic particles.
- The rise of monoclonal antibody therapy, enabled by techniques to more rapidly engineer and make human antibodies in large batches. Three monoclonal therapies have emergency use authorization for COVID-19, including GlaxoSmithKline and Vir Biotechnology’s. That therapy has as its basis a human antibody, S309, investigated by researchers at the University of Washington and other institutions.