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 . 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.
 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.
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
Posted on by Dr. Francis Collins
There’s no question that vaccines are making a tremendous difference in protecting individuals and whole communities against infection and severe illness from SARS-CoV-2, the coronavirus that causes COVID-19. And now, there’s yet another reason to get the vaccine: in the event of a breakthrough infection, people who are fully vaccinated also are substantially less likely to develop Long COVID Syndrome, which causes brain fog, muscle pain, fatigue, and a constellation of other debilitating symptoms that can last for months after recovery from an initial infection.
These important findings published in The Lancet Infectious Diseases are the latest from the COVID Symptom Study . This study allows everyday citizens in the United Kingdom to download a smartphone app and self-report data on their infection, symptoms, and vaccination status over a long period of time.
Previously, the study found that 1 in 20 people in the U.K. who got COVID-19 battled Long COVID symptoms for eight weeks or more. But this work was done before vaccines were widely available. What about the risk among those who got COVID-19 for the first time as a breakthrough infection after receiving a double dose of any of the three COVID-19 vaccines (Pfizer, Moderna, AstraZeneca) authorized for use in the U.K.?
To answer that question, Claire Steves, King’s College, London, and colleagues looked to frequent users of the COVID Symptom Study app on their smartphones. In its new work, Steves’ team was interested in analyzing data submitted by folks who’d logged their symptoms, test results, and vaccination status between December 9, 2020, and July 4, 2021. The team found there were more than 1.2 million adults who’d received a first dose of vaccine and nearly 1 million who were fully vaccinated during this period.
The data show that only 0.2 percent of those who were fully vaccinated later tested positive for COVID-19. While accounting for differences in age, sex, and other risk factors, the researchers found that fully vaccinated individuals who developed breakthrough infections were about half (49 percent) as likely as unvaccinated people to report symptoms of Long COVID Syndrome lasting at least four weeks after infection.
The most common symptoms were similar in vaccinated and unvaccinated adults with COVID-19, and included loss of smell, cough, fever, headaches, and fatigue. However, all of these symptoms were milder and less frequently reported among the vaccinated as compared to the unvaccinated.
Vaccinated people who became infected were also more likely than the unvaccinated to be asymptomatic. And, if they did develop symptoms, they were half as likely to report multiple symptoms in the first week of illness. Another vaccination benefit was that people with a breakthrough infection were about a third as likely to report any severe symptoms. They also were more than 70 percent less likely to require hospitalization.
We still have a lot to learn about Long COVID, and, to get the answers, NIH has launched the RECOVER Initiative. The initiative will study tens of thousands of COVID-19 survivors to understand why many individuals don’t recover as quickly as expected, and what might be the cause, prevention, and treatment for Long COVID.
In the meantime, these latest findings offer the encouraging news that help is already here in the form of vaccines, which provide a very effective way to protect against COVID-19 and greatly reduce the odds of Long COVID if you do get sick. So, if you haven’t done so already, make a plan to protect your own health and help end this pandemic by getting yourself fully vaccinated. Vaccines are free and available near to you—just go to vaccines.gov or text your zip code to 438829.
 Risk factors and disease profile of post-vaccination SARS-CoV-2 infection in UK users of the COVID Symptom Study app: a prospective, community-based, nested, case-control study. Antonelli M, Penfold RS, Merino J, Sudre CH, Molteni E, Berry S, Canas LS, Graham MS, Klaser K, Modat M, Murray B, Kerfoot E, Chen L, Deng J, Österdahl MF, Cheetham NJ, Drew DA, Nguyen LH, Pujol JC, Hu C, Selvachandran S, Polidori L, May A, Wolf J, Chan AT, Hammers A, Duncan EL, Spector TD, Ourselin S, Steves CJ. Lancet Infect Dis. 2021 Sep 1:S1473-3099(21)00460-6.
COVID-19 Research (NIH)
Claire Steves (King’s College London, United Kingdom)
Posted on by Dr. Francis Collins
At the end of last year, you may recall hearing news reports that the number of COVID-19 cases in the United States had topped 20 million. While that number came as truly sobering news, it also likely was an underestimate. Many cases went undetected due to limited testing early in the year and a large number of infections that produced mild or no symptoms.
Now, a recent article published in Nature offers a more-comprehensive estimate that puts the true number of infections by the end of 2020 at more than 100 million . That’s equal to just under a third of the U.S. population of 328 million. This revised number shows just how rapidly this novel coronavirus spread through the country last year. It also brings home just how timely the vaccines have been—and continue to be in 2021—to protect our nation’s health in this time of pandemic.
The work comes from NIH grantee Jeffrey Shaman, Sen Pei, and colleagues, Columbia University, New York. As shown above in the map, the researchers estimated the percentage of people who had been infected with SARS-CoV-2, the novel coronavirus that causes COVID-19, in communities across the country through December 2020.
To generate this map, they started with existing national data on the number of coronavirus cases (both detected and undetected) in 3,142 U.S. counties and major metropolitan areas. They then factored in data from the Centers for Disease Control and Prevention (CDC) on the number of people who tested positive for antibodies against SARS-CoV-2. These CDC data are useful for picking up on past infections, including those that went undetected.
From these data, the researchers calculated that only about 11 percent of all COVID-19 cases were confirmed by a positive test result in March 2020. By the end of the year, with testing improvements and heightened public awareness of COVID-19, the ascertainment rate (the number of infections that were known versus unknown) rose to about 25 percent on average. This measure also varied a lot across the country. For instance, the ascertainment rates in Miami and Phoenix were higher than the national average, while rates in New York City, Los Angeles, and Chicago were lower than average.
How many people were potentially walking around with a contagious SARS-CoV-2 infection? The model helps to answer this, too. On December 31, 2020, the researchers estimate that 0.77 percent of the U.S. population had a contagious infection. That’s about 1 in every 130 people on average. In some places, it was much higher. In Los Angeles, for example, nearly 1 in 40 (or 2.42 percent) had a SARS-CoV-2 infection as they rang in the New Year.
Over the course of the year, the fatality rate associated with COVID-19 dropped, at least in part due to earlier diagnosis and advances in treatment. The fatality rate went from 0.77 percent in April to 0.31 percent in December. While this is great news, it still shows that COVID-19 remains much more dangerous than seasonal influenza (which has a fatality rate of 0.08 percent).
Today, the landscape has changed considerably. Vaccines are now widely available, giving many more people immune protection without ever having to get infected. And yet, the rise of the Delta and other variants means that breakthrough infections and reinfections—which the researchers didn’t account for in their model—have become a much bigger concern.
Looking ahead to the end of 2021, Americans must continue to do everything they can to protect their communities from the spread of this terrible virus. That means getting vaccinated if you haven’t already, staying home and getting tested if you’ve got symptoms or know of an exposure, and taking other measures to keep yourself and your loved ones safe and well. These measures we take now will influence the infection rates and susceptibility to SARS-CoV-2 in our communities going forward. That will determine what the map of SARS-CoV-2 infections will look like in 2021 and beyond and, ultimately, how soon we can finally put this pandemic behind us.
 Burden and characteristics of COVID-19 in the United States during 2020. Pei S, Yamana TK, Kandula S, Galanti M, Shaman J. Nature. 2021 Aug 26.
COVID-19 Research (NIH)
Sen Pei (Columbia University, New York)
Jeffrey Shaman (Columbia University, New York)
Posted on by Dr. Francis Collins
Many factors influence our risk of illness from SARS-CoV-2, the coronavirus responsible for COVID-19. That includes being careful to limit our possible exposures to the virus, as well as whether we have acquired immunity from a vaccine or an earlier infection. But once a person is infected, a host of other biological factors, including age and pre-existing medical conditions, will influence one’s risk of becoming severely ill.
While earlier studies have tied COVID-19 severity to genetic variations in a person’s antiviral defenses and blood type, we still have a lot to learn about how a person’s genetic makeup influences COVID-19 susceptibility and severity. So, I was pleased to see the recent findings of an impressive global effort to map the genetic underpinnings of SARS-CoV-2 infection and COVID-19 severity, which involved analyzing the genomes of many thousands of people with COVID-19 around the globe.
This comprehensive search led to the identification of 13 regions of the human genome that appear to play a role in COVID-19 infection or severity. Though more research is needed to sort out these leads, they represent potentially high-quality clues to the pathways that this virus uses to cause illness, and help to explain why some people are more likely to become infected with SARS-CoV-2 or to develop severe disease.
The international effort, known as The COVID-19 Host Genetics Initiative, is led by Andrea Ganna, Institute for Molecular Medicine Finland, Helsinki, and colleagues in the United States and around the world. Teasing out those important genetic influences is no easy task. It requires vast amounts of data, so Ganna reached out to the scientific community via Twitter to announce a new COVID-19 gene-hunting effort and ask for help. Thousands of researchers around the world answered his call. The new study, published in the journal Nature, includes data collected through the initiative as of January 2021, and represents nearly 50,000 COVID-19 patients and another 2 million uninfected controls .
In search of common gene variants that may influence who becomes infected with SARS-CoV-2 and how sick they will become, Ganna’s international team turned to genome-wide association studies (GWAS). As part of this, the team analyzed patient genome data for millions of so-called single-nucleotide polymorphisms, or SNPs. While these single “letter” nucleotide substitutions found all across the genome are generally of no health significance, they can point the way to the locations of gene variants that turn up more often in association with particular traits or conditions—in this case, COVID-19 susceptibility or severity. To find them, the researchers compared SNPs in people with COVID-19 to those in about 2 million healthy blood donors from the same population groups. They also looked for variants that turned up significantly more often in people who became severely ill.
Their analyses uncovered a number of gene variants associated with SARS-CoV-2 infection or severe COVID-19 in 13 regions of the human genome, six of which were new. Four of the 13 affect a person’s risk for becoming infected with SARS-CoV-2. The other nine influence a person’s risk for developing severe illness following the infection.
Interestingly, some of these gene variants already were known to have associations with other types of lung or autoimmune diseases. The new findings also help to confirm previous studies suggesting that the gene that determines a person’s blood type may influence a person’s susceptibility to SARS-CoV-2 infection, along with other genes that play a role in immunity. For example, the findings show overlap with variants within a gene called TYK2, which was earlier shown to protect against autoimmune-related diseases. Some of the variants also point to the need for further work to study previously unexplored biological processes that may play potentially important roles in COVID-19.
Two of the new variants associated with disease severity were discovered only by including individuals with East Asian ancestry, highlighting the value of diversity in such analyses to gain a more comprehensive understanding of the biology. One of these newfound variants is close to a gene known as FOXP4, which is especially intriguing because this gene is known to play a role in the airways of the lung.
The researchers continue to look for more underlying clues into the biology of COVID-19. In fact, their latest unpublished analysis has increased the number of COVID-19 patients from about 50,000 to 125,000, making it possible to add another 10 gene variants to the list.
 Mapping the human genetic architecture of COVID-19. COVID-19 Host Genetics Initiative. Nature. 2021 Jul 8.
COVID-19 Research (NIH)