Proteins, proteins everywhere
PHOTO: CAMERON DAVIDSON
The first protein structures were determined by x-ray crystallography in 1957 by John C. Kendrew and Max F. Perutz. As a bioinorganic chemist, I was delighted that the structures were myoglobin and hemoglobin, both heme proteins with big, beautiful iron atoms. It must have been an extraordinary experience to stare at a physical model of the structures and see something that had previously only been imagined. Not long afterward, Christian B. Anfinsen Jr. proposed that the structure of a protein was thermodynamically stable. It seemed possible that the three-dimensional structure of a protein could be predicted based on the sequence of its amino acids. This “protein-folding problem,” as it came to be known, baffled scientists until this year, when the papers we’ve deemed the 2021 Breakthrough of the Year were published.
I came of age as a biochemist in an era when scientists could describe but not solve protein folding. Chemical force field and molecular dynamics modeling just couldn’t quite get there. The early days of de novo protein synthesis were heady because complicated structures could be designed and synthesized, but elaborate structures and multiprotein assemblies remained out of reach. When I left the lab in 2006 to go fight administrative windmills, I thought the protein-folding problem would never be solved.
A lot has happened in the 15 years since. Structures started coming much faster after the development of cryo–electron microscopy (cryo-EM), a method requiring very expensive instrumentation that can determine protein structures without crystallized samples. Then the Critical Assessment of Protein Structure Prediction (CASP) competition, in which scientists were challenged to match algorithms to known protein structures, really took off. As these annual competitions proceeded, DeepMind—a UK subsidiary of Alphabet, Google’s parent company—began showing results from an artificial intelligence algorithm that mined known protein structures to predict unknown structures. Their program, AlphaFold, made a big impression at the fall 2020 CASP conference on Minkyung Baek, a researcher at the University of Washington. Baek devised RoseTTAFold, an algorithm that uses less computing power and can predict structures of protein complexes. The methods of DeepMind and Baek et al. were published simultaneously in Nature and Science, respectively, this year. Both AlphaFold and RoseTTAFold are available free for researchers (as is a database of protein structure predictions), and scientists immediately began obtaining protein structures from these algorithms without having to crystallize their proteins or access cryo-EM tools. 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.
This is a breakthrough on two fronts. It solves a scientific problem that has been on the to-do list for 50 years. And just like Fermat’s Last Theorem or gravitational waves, scientists kept at it until it was done. Also, it’s a game-changing technique that, like CRISPR or cryo-EM, will greatly accelerate scientific discovery. It would have been the Breakthrough of the Year for either reason, but it’s a twofer for sure.
The breakdowns of the year are all topics that were covered heavily in Science. The disastrous US Food and Drug Administration (FDA) approval for Alzheimer’s disease of aducanumab—a costly treatment with little if any clinical efficacy—was discussed in an Editorial by Joel Perlmutter. Perlmutter resigned in protest from the committee that advised the FDA on the drug. Continued inaction on climate has rendered unlikely the 1.5°C target as a limit on the increase in global temperature. And scientists have been attacked in new ways and with new methods by politicians who are exploiting social media and long-tested methods of indoctrination to undermine scientific authority on issues ranging from vaccines in the midst of a pandemic to the impending devastation of climate change.
The contrasts of the breakthrough and breakdowns are stunning. The breakthrough in protein folding is one of the greatest ever in terms of both the scientific achievement and the enabling of future research. Meanwhile, the tragic loss of respect for scientific authority around the world is demoralizing, and it’s hard to see the trend reversing any time soon.
It is truly the best of times and the worst of times in 2021—an age of wisdom and an age of foolishness. The scientific tale of two cities will be with us for a while to come.
Information & Authors
Information
Published In
Science
Volume 374 | Issue 6574
17 December 2021
17 December 2021
Copyright
Copyright © 2021 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works.
Submission history
Published in print: 17 December 2021
Authors
Metrics & Citations
Metrics
Article Usage
Altmetrics
Citations
Export citation
Select the format you want to export the citation of this publication.
View Options
View options
PDF format
Download this article as a PDF file
Download PDFGet Access
Log in to view the full text
AAAS login provides access to Science for AAAS members, and access to other journals in the Science family to users who have purchased individual subscriptions.
- Become a AAAS Member
- Activate your Account
- Purchase Access to Other Journals in the Science Family
- Account Help
Log in via OpenAthens.
Log in via Shibboleth.
More options
Purchase digital access to this article
Download and print this article for your personal scholarly, research, and educational use.
Buy a single issue of Science for just $15 USD.
Will successful AI approaches yield the physical laws of protein folding?
Thirty-five years ago, when I was a graduate student in molecular biology and genetics at the Johns Hopkins University School of Medicine, some of my favorite classmates and buddies were training to become x-ray crystallographers; and several did. I was a groupie of the protein folding problem for most of my career as a cancer cell molecular biologist; studying the problem vicariously and often annoying my faculty colleagues, who were actively pursuing the folding problem, by invading their forums with my naive ideas about how proteins so efficiently achieved and maintained the dynamic structures that allow them to perform their many diverse functions with the effectiveness required for organismal life. Now learning that the protein folding problem has been breached with the high-intensity comparative computation of AI approaches is both exciting and disappointing to me. Though certainly very good, these methods are only as good as the quality of the resolved structures upon which they are based, which of course is still a huge accomplishment. But I think the next great pure science challenge for all of applied AI work is to define from their success the physical forces, factors, and laws that are responsible. Because when these are known, the quintessential solution to the protein folding problem, the ability to confidently predict the folding dynamics of proteins, based only on their amino acid sequence without comparisions to previously resolved structures, will then truly be in hand.