structural biology
Using AI to Find New Antibiotics Still a Work in Progress
Posted on by Lawrence Tabak, D.D.S., Ph.D.
Each year, more than 2.8 million people in the United States develop bacterial infections that don’t respond to treatment and sometimes turn life-threatening [1]. Their infections are antibiotic-resistant, meaning the bacteria have changed in ways that allow them to withstand our current widely used arsenal of antibiotics. It’s a serious and growing health-care problem here and around the world. To fight back, doctors desperately need new antibiotics, including novel classes of drugs that bacteria haven’t seen and developed ways to resist.
Developing new antibiotics, however, involves much time, research, and expense. It’s also fraught with false leads. That’s why some researchers have turned to harnessing the predictive power of artificial intelligence (AI) in hopes of selecting the most promising leads faster and with greater precision.
It’s a potentially paradigm-shifting development in drug discovery, and a recent NIH-funded study, published in the journal Molecular Systems Biology, demonstrates AI’s potential to streamline the process of selecting future antibiotics [2]. The results are also a bit sobering. They highlight the current limitations of one promising AI approach, showing that further refinement will still be needed to maximize its predictive capabilities.
These findings come from the lab of James Collins, Massachusetts Institute of Technology (MIT), Cambridge, and his recently launched Antibiotics-AI Project. His audacious goal is to develop seven new classes of antibiotics to treat seven of the world’s deadliest bacterial pathogens in just seven years. What makes this project so bold is that only two new classes of antibiotics have reached the market in the last 50 years!
In the latest study, Collins and his team looked to an AI program called AlphaFold2 [3]. The name might ring a bell. AlphaFold’s AI-powered ability to predict protein structures was a finalist in Science Magazine’s 2020 Breakthrough of the Year. In fact, AlphaFold has been used already to predict the structures of more than 200 million proteins, or almost every known protein on the planet [4].
AlphaFold employs a deep learning approach that can predict most protein structures from their amino acid sequences about as well as more costly and time-consuming protein-mapping techniques.
In the deep learning models used to predict protein structure, computers are “trained” on existing data. As computers “learn” to understand complex relationships within the training material, they develop a model that can then be applied for making predictions of 3D protein structures from linear amino acid sequences without relying on new experiments in the lab.
Collins and his team hoped to combine AlphaFold with computer simulations commonly used in drug discovery as a way to predict interactions between essential bacterial proteins and antibacterial compounds. If it worked, researchers could then conduct virtual rapid screens of millions of new synthetic drug compounds targeting key bacterial proteins that existing antibiotics don’t. It would also enable the rapid development of antibiotics that work in novel ways, exactly what doctors need to treat antibiotic-resistant infections.
To test the strategy, Collins and his team focused first on the predicted structures of 296 essential proteins from the Escherichia coli bacterium as well as 218 antibacterial compounds. Their computer simulations then predicted how strongly any two molecules (essential protein and antibacterial) would bind together based on their shapes and physical properties.
It turned out that screening many antibacterial compounds against many potential targets in E. coli led to inaccurate predictions. For example, when comparing their computational predictions with actual interactions for 12 essential proteins measured in the lab, they found that their simulated model had about a 50:50 chance of being right. In other words, it couldn’t identify true interactions between drugs and proteins any better than random guessing.
They suspect one reason for their model’s poor performance is that the protein structures used to train the computer are fixed, not flexible and shifting physical configurations as happens in real life. To improve their success rate, they ran their predictions through additional machine-learning models that had been trained on data to help them “learn” how proteins and other molecules reconfigure themselves and interact. While this souped-up model got somewhat better results, the researchers report that they still aren’t good enough to identify promising new drugs and their protein targets.
What now? In future studies, the Collins lab will continue to incorporate and train the computers on even more biochemical and biophysical data to help with the predictive process. That’s why this study should be interpreted as an interim progress report on an area of science that will only get better with time.
But it’s also a sobering reminder that the quest to find new classes of antibiotics won’t be easy—even when aided by powerful AI approaches. We certainly aren’t there yet, but I’m confident that we will get there to give doctors new therapeutic weapons and turn back the rise in antibiotic-resistant infections.
References:
[1] 2019 Antibiotic resistance threats report. Centers for Disease Control and Prevention.
[2] Benchmarking AlphaFold-enabled molecular docking predictions for antibiotic discovery. Wong F, Krishnan A, Zheng EJ, Stark H, Manson AL, Earl AM, Jaakkola T, Collins JJ. Molecular Systems Biology. 2022 Sept 6. 18: e11081.
[3] Highly accurate protein structure prediction with AlphaFold. Jumper J, Evans R, Pritzel A, Kavukcuoglu K, Kohli P, Hassabis D., et al. Nature. 2021 Aug;596(7873):583-589.
[4] ‘The entire protein universe’: AI predicts shape of nearly every known protein. Callaway E. Nature. 2022 Aug;608(7921):15-16.
Links:
Antimicrobial (Drug) Resistance (National Institute of Allergy and Infectious Diseases/NIH)
Collins Lab (Massachusetts Institute of Technology, Cambridge)
The Antibiotics-AI Project, The Audacious Project (TED)
AlphaFold (Deep Mind, London, United Kingdom)
NIH Support: National Institute of Allergy and Infectious Diseases; National Institute of General Medical Sciences
The Amazing Brain: Capturing Neurons in Action
Posted on by Lawrence Tabak, D.D.S., Ph.D.
With today’s powerful imaging tools, neuroscientists can monitor the firing and function of many distinct neurons in our brains, even while we move freely about. They also possess another set of tools to capture remarkable, high-resolution images of the brain’s many thousands of individual neurons, tracing the form of each intricate branch of their tree-like structures.
Most brain imaging approaches don’t capture neural form and function at once. Yet that’s precisely what you’re seeing in this knockout of a movie, another winner in the Show Us Your BRAINs! Photo and Video Contest, supported by NIH’s Brain Research through Advancing Innovative Neurotechnologies® (BRAIN) Initiative.
This first-of-its kind look into the mammalian brain produced by Andreas Tolias, Baylor College of Medicine, Houston, and colleagues features about 200 neurons in the visual cortex, which receives and processes visual information. First, you see a colorful, tightly packed network of neurons. Then, those neurons, which were colorized by the researchers in vibrant pinks, reds, blues, and greens, pull apart to reveal their finely detailed patterns and shapes. Throughout the video, you can see neural activity, which appears as flashes of white that resemble lightning bolts.
Making this movie was a multi-step process. First, the Tolias group presented laboratory mice with a series of visual cues, using a functional imaging approach called two-photon calcium imaging to record the electrical activity of individual neurons. While this technique allowed the researchers to pinpoint the precise locations and activity of each individual neuron in the visual cortex, they couldn’t zoom in to see their precise structures.
So, the Baylor team sent the mice to colleagues Nuno da Costa and Clay Reid, Allen Institute for Brain Science, Seattle, who had the needed electron microscopes and technical expertise to zoom in on these structures. Their data allowed collaborator Sebastian Seung’s team, Princeton University, Princeton, NJ, to trace individual neurons in the visual cortex along their circuitous paths. Finally, they used sophisticated machine learning algorithms to carefully align the two imaging datasets and produce this amazing movie.
This research was supported by Intelligence Advanced Research Projects Activity (IARPA), part of the Office of the Director of National Intelligence. The IARPA is one of NIH’s governmental collaborators in the BRAIN Initiative.
Tolias and team already are making use of their imaging data to learn more about the precise ways in which individual neurons and groups of neurons in the mouse visual cortex integrate visual inputs to produce a coherent view of the animals’ surroundings. They’ve also collected an even-larger data set, scaling their approach up to tens of thousands of neurons. Those data are now freely available to other neuroscientists to help advance their work. As researchers make use of these and similar data, this union of neural form and function will surely yield new high-resolution discoveries about the mammalian brain.
Links:
Tolias Lab (Baylor College of Medicine, Houston)
Nuno da Costa (Allen Institute for Brain Science, Seattle)
R. Clay Reid (Allen Institute)
H. Sebastian Seung (Princeton University, Princeton, NJ)
Machine Intelligence from Cortical Networks (MICrONS) Explorer
Brain Research through Advancing Innovative Neurotechnologies® (BRAIN) Initiative (NIH)
Show Us Your BRAINs Photo & Video Contest (BRAIN Initiative)
NIH Support: BRAIN Initiative; Common Fund
Biomedical Research Leads Science’s 2021 Breakthroughs
Posted on by Lawrence Tabak, D.D.S., Ph.D.
Hi everyone, I’m Larry Tabak. I’ve served as NIH’s Principal Deputy Director for over 11 years, and I will be the acting NIH director until a new permanent director is named. In my new role, my day-to-day responsibilities will certainly increase, but I promise to carve out time to blog about some of the latest research progress on COVID-19 and any other areas of science that catch my eye.
I’ve also invited the directors of NIH’s Institutes and Centers (ICs) to join me in the blogosphere and write about some of the cool science in their research portfolios. I will publish a couple of posts to start, then turn the blog over to our first IC director. From there, I envision alternating between posts from me and from various IC directors. That way, we’ll cover a broad array of NIH science and the tremendous opportunities now being pursued in biomedical research.
Since I’m up first, let’s start where the NIH Director’s Blog usually begins each year: by taking a look back at Science’s Breakthroughs of 2021. The breakthroughs were formally announced in December near the height of the holiday bustle. In case you missed the announcement, the biomedical sciences accounted for six of the journal Science’s 10 breakthroughs. Here, I’ll focus on four biomedical breakthroughs, the ones that NIH has played some role in advancing, starting with Science’s editorial and People’s Choice top-prize winner:
Breakthrough of the Year: AI-Powered Predictions of Protein Structure
The biochemist Christian Anfinsen, who had a distinguished career at NIH, shared the 1972 Nobel Prize in Chemistry, for work suggesting that the biochemical interactions among the amino acid building blocks of proteins were responsible for pulling them into the final shapes that are essential to their functions. In his Nobel acceptance speech, Anfinsen also made a bold prediction: one day it would be possible to determine the three-dimensional structure of any protein based on its amino acid sequence alone. Now, with advances in applying artificial intelligence to solve biological problems—Anfinsen’s bold prediction has been realized.
But getting there wasn’t easy. Every two years since 1994, research teams from around the world have gathered to compete against each other in developing computational methods for predicting protein structures from sequences alone. A score of 90 or above means that a predicted structure is extremely close to what’s known from more time-consuming work in the lab. In the early days, teams more often finished under 60.
In 2020, a London-based company called DeepMind made a leap with their entry called AlphaFold. Their deep learning approach—which took advantage of 170,000 proteins with known structures—most often scored above 90, meaning it could solve most protein structures about as well as more time-consuming and costly experimental protein-mapping techniques. (AlphaFold was one of Science’s runner-up breakthroughs last year.)
This year, the NIH-funded lab of David Baker and Minkyung Baek, University of Washington, Seattle, Institute for Protein Design, published that their artificial intelligence approach, dubbed RoseTTAFold, could accurately predict 3D protein structures from amino acid sequences with only a fraction of the computational processing power and time that AlphaFold required [1]. They immediately applied it to solve hundreds of new protein structures, including many poorly known human proteins with important implications for human health.
The DeepMind and RoseTTAFold scientists continue to solve more and more proteins [1,2], both alone and in complex with other proteins. The code is now freely available for use by researchers anywhere in the world. In one timely example, AlphaFold helped to predict the structural changes in spike proteins of SARS-CoV-2 variants Delta and Omicron [3]. This ability to predict protein structures, first envisioned all those years ago, now promises to speed fundamental new discoveries and the development of new ways to treat and prevent any number of diseases, making it this year’s Breakthrough of the Year.
Anti-Viral Pills for COVID-19
The development of the first vaccines to protect against COVID-19 topped Science’s 2020 breakthroughs. This year, we’ve also seen important progress in treating COVID-19, including the development of anti-viral pills.
First, there was the announcement in October of interim data from Merck, Kenilworth, NJ, and Ridgeback Biotherapeutics, Miami, FL, of a significant reduction in hospitalizations for those taking the anti-viral drug molnupiravir [4] (originally developed with an NIH grant to Emory University, Atlanta). Soon after came reports of a Pfizer anti-viral pill that might target SARS-CoV-2, the novel coronavirus that causes COVID-19, even more effectively. Trial results show that, when taken within three days of developing COVID-19 symptoms, the pill reduced the risk of hospitalization or death in adults at high risk of progressing to severe illness by 89 percent [5].
On December 22, the Food and Drug Administration (FDA) granted Emergency Use Authorization (EUA) for Pfizer’s Paxlovid to treat mild-to-moderate COVID-19 in people age 12 and up at high risk for progressing to severe illness, making it the first available pill to treat COVID-19 [6]. The following day, the FDA granted an EUA for Merck’s molnupiravir to treat mild-to-moderate COVID-19 in unvaccinated, high-risk adults for whom other treatment options aren’t accessible or recommended, based on a final analysis showing a 30 percent reduction in hospitalization or death [7].
Additional promising anti-viral pills for COVID-19 are currently in development. For example, a recent NIH-funded preclinical study suggests that a drug related to molnupiravir, known as 4’-fluorouridine, might serve as a broad spectrum anti-viral with potential to treat infections with SARS-CoV-2 as well as respiratory syncytial virus (RSV) [8].
Artificial Antibody Therapies
Before anti-viral pills came on the scene, there’d been progress in treating COVID-19, including the development of monoclonal antibody infusions. Three monoclonal antibodies now have received an EUA for treating mild-to-moderate COVID-19, though not all are effective against the Omicron variant [9]. This is also an area in which NIH’s Accelerating COVID-19 Therapeutic Interventions and Vaccines (ACTIV) public-private partnership has made big contributions.
Monoclonal antibodies are artificially produced versions of the most powerful antibodies found in animal or human immune systems, made in large quantities for therapeutic use in the lab. Until recently, this approach had primarily been put to work in the fight against conditions including cancer, asthma, and autoimmune diseases. That changed in 2021 with success using monoclonal antibodies against infections with SARS-CoV-2 as well as respiratory syncytial virus (RSV), human immunodeficiency virus (HIV), and other infectious diseases. This earned them a prominent spot among Science’s breakthroughs of 2021.
Monoclonal antibodies delivered via intravenous infusions continue to play an important role in saving lives during the pandemic. But, there’s still room for improvement, including new formulations highlighted on the blog last year that might be much easier to deliver.
CRISPR Fixes Genes Inside the Body
One of the most promising areas of research in recent years has been gene editing, including CRISPR/Cas9, for fixing misspellings in genes to treat or even cure many conditions. This year has certainly been no exception.
CRISPR is a highly precise gene-editing system that uses guide RNA molecules to direct a scissor-like Cas9 enzyme to just the right spot in the genome to cut out or correct disease-causing misspellings. Science highlights a small study reported in The New England Journal of Medicine by researchers at Intellia Therapeutics, Cambridge, MA, and Regeneron Pharmaceuticals, Tarrytown, NY, in which six people with hereditary transthyretin (TTR) amyloidosis, a condition in which TTR proteins build up and damage the heart and nerves, received an infusion of guide RNA and CRISPR RNA encased in tiny balls of fat [10]. The goal was for the liver to take them up, allowing Cas9 to cut and disable the TTR gene. Four weeks later, blood levels of TTR had dropped by at least half.
In another study not yet published, researchers at Editas Medicine, Cambridge, MA, injected a benign virus carrying a CRISPR gene-editing system into the eyes of six people with an inherited vision disorder called Leber congenital amaurosis 10. The goal was to remove extra DNA responsible for disrupting a critical gene expressed in the eye. A few months later, two of the six patients could sense more light, enabling one of them to navigate a dimly lit obstacle course [11]. This work builds on earlier gene transfer studies begun more than a decade ago at NIH’s National Eye Institute.
Last year, in a research collaboration that included former NIH Director Francis Collins’s lab at the National Human Genome Research Institute (NHGRI), we also saw encouraging early evidence in mice that another type of gene editing, called DNA base editing, might one day correct Hutchinson-Gilford Progeria Syndrome, a rare genetic condition that causes rapid premature aging. Preclinical work has even suggested that gene-editing tools might help deliver long-lasting pain relief. The technology keeps getting better, too. This isn’t the first time that gene-editing advances have landed on Science’s annual Breakthrough of the Year list, and it surely won’t be the last.
The year 2021 was a difficult one as the pandemic continued in the U.S. and across the globe, taking far too many lives far too soon. But through it all, science has been relentless in seeking and finding life-saving answers, from the rapid development of highly effective COVID-19 vaccines to the breakthroughs highlighted above.
As this list also attests, the search for answers has progressed impressively in other research areas during these difficult times. These groundbreaking discoveries are something in which we can all take pride—even as they encourage us to look forward to even bigger breakthroughs in 2022. Happy New Year!
References:
[1] Accurate prediction of protein structures and interactions using a three-track neural network. Baek M, DiMaio F, Anishchenko I, Dauparas J, Grishin NV, Adams PD, Read RJ, Baker D., et al. Science. 2021 Jul 15:eabj8754.
[2] Highly accurate protein structure prediction with AlphaFold. Jumper J, Evans R, Pritzel A, Green T, Senior AW, Kavukcuoglu K, Kohli P, Hassabis D. et al. Nature. 2021 Jul 15.
[3] Structural insights of SARS-CoV-2 spike protein from Delta and Omicron variants. Sadek A, Zaha D, Ahmed MS. preprint bioRxiv. 2021 Dec 9.
[5] Pfizer’s novel COVID-19 oral antiviral treatment candidate reduced risk of hospitalization or death by 89% in interim analysis of phase 2/3 EPIC-HR Study. Pfizer. 5 November 52021.
[6] Coronavirus (COVID-19) Update: FDA authorizes first oral antiviral for treatment of COVID-19. Food and Drug Administration. 22 Dec 2021.
[7] Coronavirus (COVID-19) Update: FDA authorizes additional oral antiviral for treatment of COVID-19 in certain adults. Food and Drug Administration. 23 Dec 2021.
[8] 4′-Fluorouridine is an oral antiviral that blocks respiratory syncytial virus and SARS-CoV-2 replication. Sourimant J, Lieber CM, Aggarwal M, Cox RM, Wolf JD, Yoon JJ, Toots M, Ye C, Sticher Z, Kolykhalov AA, Martinez-Sobrido L, Bluemling GR, Natchus MG, Painter GR, Plemper RK. Science. 2021 Dec 2.
[9] Anti-SARS-CoV-2 monoclonal antibodies. NIH COVID-19 Treatment Guidelines. 16 Dec 2021.
[10] CRISPR-Cas9 in vivo gene editing for transthyretin amyloidosis. Gillmore JD, Gane E, Taubel J, Kao J, Fontana M, Maitland ML, Seitzer J, O’Connell D, Walsh KR, Wood K, Phillips J, Xu Y, Amaral A, Boyd AP, Cehelsky JE, McKee MD, Schiermeier A, Harari O, Murphy A, Kyratsous CA, Zambrowicz B, Soltys R, Gutstein DE, Leonard J, Sepp-Lorenzino L, Lebwohl D. N Engl J Med. 2021 Aug 5;385(6):493-502.
[11] Editas Medicine announces positive initial clinical data from ongoing phase 1/2 BRILLIANCE clinical trial of EDIT-101 For LCA10. Editas Medicine. 29 Sept 2021.
Links:
Structural Biology (National Institute of General Medical Sciences/NIH)
The Structures of Life (NIGMS)
COVID-19 Research (NIH)
2021 Science Breakthrough of the Year (American Association for the Advancement of Science, Washington, D.C)
Artificial Intelligence Accurately Predicts RNA Structures, Too
Posted on by Dr. Francis Collins
Researchers recently showed that a computer could “learn” from many examples of protein folding to predict the 3D structure of proteins with great speed and precision. Now a recent study in the journal Science shows that a computer also can predict the 3D shapes of RNA molecules [1]. This includes the mRNA that codes for proteins and the non-coding RNA that performs a range of cellular functions.
This work marks an important basic science advance. RNA therapeutics—from COVID-19 vaccines to cancer drugs—have already benefited millions of people and will help many more in the future. Now, the ability to predict RNA shapes quickly and accurately on a computer will help to accelerate understanding these critical molecules and expand their healthcare uses.
Like proteins, the shapes of single-stranded RNA molecules are important for their ability to function properly inside cells. Yet far less is known about these RNA structures and the rules that determine their precise shapes. The RNA elements (bases) can form internal hydrogen-bonded pairs, but the number of possible combinations of pairings is almost astronomical for any RNA molecule with more than a few dozen bases.
In hopes of moving the field forward, a team led by Stephan Eismann and Raphael Townshend in the lab of Ron Dror, Stanford University, Palo Alto, CA, looked to a machine learning approach known as deep learning. It is inspired by how our own brain’s neural networks process information, learning to focus on some details but not others.
In deep learning, computers look for patterns in data. As they begin to “see” complex relationships, some connections in the network are strengthened while others are weakened.
One of the things that makes deep learning so powerful is it doesn’t rely on any preconceived notions. It also can pick up on important features and patterns that humans can’t possibly detect. But, as successful as this approach has been in solving many different kinds of problems, it has primarily been applied to areas of biology, such as protein folding, in which lots of data were available for researchers to train the computers.
That’s not the case with RNA molecules. To work around this problem, Dror’s team designed a neural network they call ARES. (No, it’s not the Greek god of war. It’s short for Atomic Rotationally Equivariant Scorer.)
To start, the researchers trained ARES on just 18 small RNA molecules for which structures had been experimentally determined. They gave ARES these structural models specified only by their atomic structure and chemical elements.
The next test was to see if ARES could determine from this small training set the best structural model for RNA sequences it had never seen before. The researchers put it to the test with RNA molecules whose structures had been determined more recently.
ARES, however, doesn’t come up with the structures itself. Instead, the researchers give ARES a sequence and at least 1,500 possible 3D structures it might take, all generated using another computer program. Based on patterns in the training set, ARES scores each of the possible structures to find the one it predicts is closest to the actual structure. Remarkably, it does this without being provided any prior information about features important for determining RNA shapes, such as nucleotides, steric constraints, and hydrogen bonds.
It turns out that ARES consistently outperforms humans and all other previous methods to produce the best results. In fact, it outperformed at least nine other methods to come out on top in a community-wide RNA-puzzles contest. It also can make predictions about RNA molecules that are significantly larger and more complex than those upon which it was trained.
The success of ARES and this deep learning approach will help to elucidate RNA molecules with potentially important implications for health and disease. It’s another compelling example of how deep learning promises to solve many other problems in structural biology, chemistry, and the material sciences when—at the outset—very little is known.
Reference:
[1] Geometric deep learning of RNA structure. Townshend RJL, Eismann S, Watkins AM, Rangan R, Karelina M, Das R, Dror RO. Science. 2021 Aug 27;373(6558):1047-1051.
Links:
Structural Biology (National Institute of General Medical Sciences/NIH)
The Structures of Life (National Institute of General Medical Sciences/NIH)
RNA Biology (NIH)
Dror Lab (Stanford University, Palo Alto, CA)
NIH Support: National Cancer Institute; National Institute of General Medical Sciences
Artificial Intelligence Accurately Predicts Protein Folding
Posted on by Dr. Francis Collins
Proteins are the workhorses of the cell. Mapping the precise shapes of the most important of these workhorses helps to unlock their life-supporting functions or, in the case of disease, potential for dysfunction. While the amino acid sequence of a protein provides the basis for its 3D structure, deducing the atom-by-atom map from principles of quantum mechanics has been beyond the ability of computer programs—until now.
In a recent study in the journal Science, researchers reported they have developed artificial intelligence approaches for predicting the three-dimensional structure of proteins in record time, based solely on their one-dimensional amino acid sequences [1]. This groundbreaking approach will not only aid researchers in the lab, but guide drug developers in coming up with safer and more effective ways to treat and prevent disease.
This new NIH-supported advance is now freely available to scientists around the world. In fact, it has already helped to solve especially challenging protein structures in cases where experimental data were lacking and other modeling methods hadn’t been enough to get a final answer. It also can now provide key structural information about proteins for which more time-consuming and costly imaging data are not yet available.
The new work comes from a group led by David Baker and Minkyung Baek, University of Washington, Seattle, Institute for Protein Design. Over the course of the pandemic, Baker’s team has been working hard to design promising COVID-19 therapeutics. They’ve also been working to design proteins that might offer promising new ways to treat cancer and other conditions. As part of this effort, they’ve developed new computational approaches for determining precisely how a chain of amino acids, which are the building blocks of proteins, will fold up in space to form a finished protein.
But the ability to predict a protein’s precise structure or shape from its sequence alone had proven to be a difficult problem to solve despite decades of effort. In search of a solution, research teams from around the world have come together every two years since 1994 at the Critical Assessment of Structure Prediction (CASP) meetings. At these gatherings, teams compete against each other with the goal of developing computational methods and software capable of predicting any of nature’s 200 million or more protein structures from sequences alone with the greatest accuracy.
Last year, a London-based company called DeepMind shook up the structural biology world with their entry into CASP called AlphaFold. (AlphaFold was one of Science’s 2020 Breakthroughs of the Year.) They showed that their artificial intelligence approach—which took advantage of the 170,000 proteins with known structures in a reiterative process called deep learning—could predict protein structure with amazing accuracy. In fact, it could predict most protein structures almost as accurately as other high-resolution protein mapping techniques, including today’s go-to strategies of X-ray crystallography and cryo-EM.
The DeepMind performance showed what was possible, but because the advances were made by a world-leading deep learning company, the details on how it worked weren’t made publicly available at the time. The findings left Baker, Baek, and others eager to learn more and to see if they could replicate the impressive predictive ability of AlphaFold outside of such a well-resourced company.
In the new work, Baker and Baek’s team has made stunning progress—using only a fraction of the computational processing power and time required by AlphaFold. The new software, called RoseTTAFold, also relies on a deep learning approach. In deep learning, computers look for patterns in large collections of data. As they begin to recognize complex relationships, some connections in the network are strengthened while others are weakened. The finished network is typically composed of multiple information-processing layers, which operate on the data to return a result—in this case, a protein structure.
Given the complexity of the problem, instead of using a single neural network, RoseTTAFold relies on three. The three-track neural network integrates and simultaneously processes one-dimensional protein sequence information, two-dimensional information about the distance between amino acids, and three-dimensional atomic structure all at once. Information from these separate tracks flows back and forth to generate accurate models of proteins rapidly from sequence information alone, including structures in complex with other proteins.
As soon as the researchers had what they thought was a reasonable working approach to solve protein structures, they began sharing it with their structural biologist colleagues. In many cases, it became immediately clear that RoseTTAFold worked remarkably well. What’s more, it has been put to work to solve challenging structural biology problems that had vexed scientists for many years with earlier methods.
RoseTTAFold already has solved hundreds of new protein structures, many of which represent poorly understood human proteins. The 3D rendering of a complex showing a human protein called interleukin-12 in complex with its receptor (above image) is just one example. The researchers have generated other structures directly relevant to human health, including some that are related to lipid metabolism, inflammatory conditions, and cancer. The program is now available on the web and has been downloaded by dozens of research teams around the world.
Cryo-EM and other experimental mapping methods will remain essential to solve protein structures in the lab. But with the artificial intelligence advances demonstrated by RoseTTAFold and AlphaFold, which has now also been released in an open-source version and reported in the journal Nature [2], researchers now can make the critical protein structure predictions at their desktops. This newfound ability will be a boon to basic science studies and has great potential to speed life-saving therapeutic advances.
References:
[1] Accurate prediction of protein structures and interactions using a three-track neural network. Baek M, DiMaio F, Anishchenko I, Dauparas J, Grishin NV, Adams PD, Read RJ, Baker D., et al. Science. 2021 Jul 15:eabj8754.
[2] Highly accurate protein structure prediction with AlphaFold. Jumper J, Evans R, Pritzel A, Green T, Senior AW, Kavukcuoglu K, Kohli P, Hassabis D. et al. Nature. 2021 Jul 15.
Links:
Structural Biology (National Institute of General Medical Sciences/NIH)
The Structures of Life (NIGMS)
Baker Lab (University of Washington, Seattle)
CASP 14 (University of California, Davis)
NIH Support: National Institute of Allergy and Infectious Diseases; National Institute of General Medical Sciences
Next Page
