In 2009, USAID, the US government agency responsible for international development, initiated Predict, a groundbreaking project for its time.
The $200 million program was tasked with building other countries’ capabilities to detect new viruses and manage outbreaks, studying the human-wildlife interface and learning about how viruses cross over into humans. Its headline work was viral discovery: directly finding novel viruses in wildlife that posed a risk of a pandemic before they spilled over, and ideally, prevent it from happening.
But after a little more than a decade, the program’s funding was cut off by the Trump administration in October 2019 — right before the novel coronavirus hit. (At the time, I mourned its passing and saw it as one more sign that we weren’t ready for the next pandemic.) And in 2021, with the threat of pandemics firmly established by the catastrophic impact of Covid-19, the Biden administration announced plans to restart a viral discovery program, this time under a new name: Deep Vzn.
Deep Vzn is an acronym for Discovery and Exploration of Emerging Pathogens — Viral Zoonoses. It’s a five-year, $125 million endeavor to send out teams all over the world to identify potentially dangerous pathogens in the wild, bring those viruses back to the lab, and perform experiments to identify which ones could seed the next pandemic.
Some of that work happened under Predict’s umbrella as well, but even at the time of the earlier program, some virologists were quietly saying that viral discovery was overhyped and a waste of time. And in the years since Predict launched, the conversation on the value of viral discovery has shifted toward even greater pessimism.
Critics — including researchers who study biosecurity and biosafety — argue it doesn’t really pass a cost-benefit analysis. In some ways, virus hunting is looking for a needle in a haystack — the handful of viruses that might cross over to humans amid tens of thousands that won’t — when we don’t even know how to tell needles from hay, or what to do with a needle once we identify one.
And some experts are raising another, even sharper question: What if viral discovery is not just an ineffective tactic but a terrible idea, one that might not only fail to prevent the next pandemic but potentially even make it more likely?
“Do you really want to be going into these bat caves to collect and then catalogue which ones are most dangerous to humans?” Andy Weber, assistant secretary of defense for nuclear, chemical, and biological defense programs under the Obama administration, told me.
His concern isn’t just that we’re looking for a needle in a haystack that we may never find. It’s that if we did discover a virus that would devastate the world if it crossed over into humans, someone might expose themselves accidentally while researching it, as has happened with smallpox and with influenzas. Worse, finding a virus and infecting animals with it in a lab could open the door to accidental release or intentional use. Success, in other words, could be worse than failure.
Monitoring the interface between humans and animals for pandemic prevention has value, particularly when the programs are narrowly targeted at certain objectives: say, a focus on reducing spillover, or surveillance of potential animal infections, or studying viruses that have already spilled over into humans. Research published last month in Nature projects that global warming could drive 4,000 viruses to spread for the first time between mammals, including potentially humans and animals, by 2070, underscoring the changing threat from zoonotic spillovers.
But if the risks of virus hunting are higher than the odds of a virus crossing over into humans and sparking a pandemic naturally, then viral discovery doesn’t just look inefficient. It looks like a bad idea.
Finding viruses in the deepest reaches of the natural world
The concept of viral discovery is simple: Every disease that might cause a naturally occurring pandemic is out there somewhere in the environment. What if we found it before it found us?
That was the concept behind one plank of Predict’s work, which sampled “at least 931 novel virus species from 145,000 samples of wildlife, livestock, and humans,” according to a 2020 paper by biologist Colin Carlson, of Georgetown University.
The Global Virome Project has a similar aim. Launched in 2018 and estimated to cost between $1 billion and $4 billion, it aims to go out into the wild and test animals for viruses. (The Global Virome Project did not reply to a request for comment.)
The key idea behind Predict, the Global Virome Project, and Deep Vzn was that if we build a catalog of hundreds of thousands of viruses out there in nature, we will figure out which ones threaten humans, and then we’ll be better prepared if and when they spill over.
“Developing these tools now is essential for being better prepared for the future when new viruses spillover and stopping them from causing outbreaks that could become pandemics,” USAID’s announcement of Deep Vzn as a program declared.
The idea behind these initiatives makes intuitive sense. The notion of being proactive in searching for the next deadly virus that could hobble humanity certainly holds appeal, especially in the post-Covid age. But scientifically, the rationale for such a program rests on more questionable ground than many of its backers assume, according to some experts.
“I still fail to see at this point how it’s going to better prepare the human race for the next infectious disease that jumps from animals to humans,” Michael Osterholm, the director of the University of Minnesota’s Center for Infectious Disease Research and Policy, has argued.
“Before the pandemic, the dominant paradigm was that if we could find these threats we could predict and prevent [the next pandemic],” Carlson, the biologist, told me. “It was a silly thing to believe even without the pandemic. ... There has been a disconnect between the proposed benefits and the reality for a while.”
Carlson’s paper goes further. “History tells us viral discovery is not enough to prevent pandemics: influenza was first isolated in 1933, Zika in 1947, chikungunya in 1952, and amid the emergence of severe acute respiratory syndrome coronavirus in 2003 and Middle East respiratory syndrome coronavirus in 2012, nearly two decades of wildlife sampling has turned up hundreds of new coronavirus species,” he writes.
And yet, knowing that a virus exists among thousands of viruses in nature, it appears, doesn’t by itself do much of anything to help us defend against it. Supporters argue that it can help with developing vaccines or treatments. But there’s yet to be an example of a successful human vaccine or treatment development program for a virus identified only in the wild — and most of the time it takes to get vaccines or treatments ready for wide-scale use is spent on clinical trials in humans that are not conducted for viruses that have been found only in animals.
It’s hard to rule out that any particular avenue of scientific research might turn up an important insight down the line. Not much has come of the viral discovery elements of Predict (more on that below), but there’s always the chance something is just around the corner to be discovered.
But with all that said, many prominent researchers remain skeptical. “Broad genomic surveys of animal viruses will ... be of little practical value when it comes to understanding and mitigating the emergence of disease,” leading virologists Edward Holmes, Andrew Rambaut, and Kristian Andersen argued in Nature in 2018, in a commentary titled “Pandemics: spend on surveillance, not prediction.” “We urge those working on infectious disease to focus funds and efforts on a much simpler and more cost-effective way to mitigate outbreaks — proactive, real-time surveillance of human populations.”
And there’s an even worse risk here to ponder.
Could viral discovery risk causing the pandemics it’s meant to stop?
In general, much pandemic prevention work has focused on minimizing human-animal interfaces — for example, encouraging people not to hunt and eat animals that are disease reservoirs, and not to go in caves full of disease-carrying bats.
But virus hunting itself frequently involves exposure to the highest-risk human-animal interfaces. One 2021 Science article about Predict tells an anecdote about the virology researchers trying to find new viruses in the Amazon: “Monkeys have bitten and sneezed on Gordo [a virus hunter profiled], and on this trip a syringe broke as he squeezed the plunger, spraying monkey blood on his face shield. He says his wife complains when he stashes monkey carcasses in their home fridge.” The tone is lighthearted, but the content is, considered from the perspective of closely working with potentially dangerous viruses, fairly terrifying.
In China, a researcher looking for bat coronaviruses “once forgot personal protective equipment and was splattered with bat urine, leading him to quarantine at home for two weeks. On multiple occasions, bat blood squirted onto his skin while he was trying to grasp the animals with a clamp,” the Washington Post reported, citing interviews in Chinese state media.
Research under those conditions might find previously undiscovered viruses. It also might spread them.
“USAID takes biosafety and biosecurity extremely seriously and has established detailed safety protocols and procedures to ensure this work is done safely,” a USAID spokesperson told me, though they did not share details.
Another concern is what happens once viruses are taken to the lab for testing and characterization, which often involves infecting lab animals with the virus to see whether and how they’re affected.
“They want to take the viruses that look the scariest, and take them back to the lab, and do experiments on them to determine which really pose a threat of a pandemic,” said Kevin Esvelt, a biologist at MIT known for his pioneering work on the gene-editing technology CRISPR. “As soon as you take them back to the lab and start working with them, you run the risk of accidental pandemics” — for example, from lab escapes, where a virus under controlled conditions makes it out of the lab and into the general population.
But that’s not even the most significant risk from such research, Weber says. “The biggest concern is that in the process of identifying potential pandemic pathogens we are actually giving a cookbook to potential bad actors,” he warns.
His argument: Let’s say you are a state actor starting a bioweapons program, or a terrorist group like the Aum Shinrikyo cult, which in the 1990s actively tried to build biological agents it could use to harm civilians.
Wouldn’t a public, neatly ordered list of genomes for all the most dangerous viruses humanity has been able to identify — for which there is no natural immunity and no stockpiled vaccines — provide the perfect shopping list?
“Once a pandemic-capable pathogen has been identified, its genome features high dual-use potential: it may inform biosurveillance while also constituting a blueprint to cause widespread harm,” a recent preprint paper from researchers at Oxford and Georgetown concluded.
Part of USAID’s plan for Deep Vzn is that all of the discovered genomes would be fully public, which is itself a response to legitimate previous concerns that viral discovery work involved the US going into poor countries and collecting data that the US then didn’t share with locals.
Esvelt puts it like this: “As soon as we publicly identify pandemic-capable viruses, we’ll be giving tens of thousands of individuals the ability to kill as many people as a nuclear device could.” In other words, knowing in advance that a virus might spill over and kill millions of people would theoretically be great. But if scientists effectively tell the world “this virus, if it infected humans, would kill millions of people,” then they’ve created a clear information hazard, accidentally opening the door to potential cataclysmic harm.
Developing effective bioweapons is difficult — but the hard part isn’t the doing, it’s identifying the rare one that is contagious and dangerous to human beings. If well-intentioned research does that part and a list of such viruses is published, then weaponizing them is quite doable even for a small team. “My own skills are rusty but I could probably do it myself,” Esvelt told me.
“The way the life sciences work is that they post the DNA of everything publicly,” Weber told me. “That’s inevitably going to enable bad actors. The sequences are the recipes for the world’s most dangerous weapons.”
How much of a threat is that, really? Don’t we already have deadly diseases? Sure, terrorists could build a pandemic virus identified through Deep Vzn, but couldn’t they also build smallpox or the 1918 flu? (The genomes for both are available.)
“We live in an era where people can create viruses if they have the blueprint,” Carlson told me. But he’s not worried that virus hunting could add new blueprints: “I believe that in terms of containment scenarios a flu is a bigger fear, and we certainly don’t say that all flu sequences should be confidential. The marginal risk is very small.”
Esvelt disagrees. “The key point to get across ... is that right now we don’t actually know of any pandemic-capable viruses” that spread in humans for which a vaccine doesn’t exist, he told me. There’s smallpox, but the US has hundreds of millions of vaccine doses on hand (and for complicated technical reasons, poxes are harder to create from a blueprint in a lab than flus or coronaviruses are, though not impossible). There are influenzas that have already hit human populations, for which we also have vaccines (and some natural immunity).
“We are partly protected by our limited knowledge of specific genotypes, mechanisms, and other critical biological details” of how best to kick off a deadly pandemic, the Oxford/Georgetown paper finds.
Is identifying new recipes for mass death worth it? That comes down to a crucial question: Does having such recipes aid in “defense” against pandemics more than it aids in “offense”?
A look at viral discovery’s track record
The case for work like Deep Vzn’s viral discovery is simple: What if scientists had known in advance that Covid-19 was circulating in wild animals, and had known it posed a threat to humans?
In that case, they could have gotten a head start on developing vaccines and treatments. If the next Covid-19 is identified while it’s still in animal hosts, the world could potentially prevent it from spilling over — or least be ready for it if it does by designing broad-spectrum vaccines and treatments.
The problem is that the world did take exactly this approach to identifying risky coronaviruses after SARS outbreaks in the early 2000s. US programs like Predict funded research to collect pathogens in the wild, including partnerships with the Wuhan Institute of Virology to collect and study coronaviruses — partnerships that hit the headlines when the coronavirus pandemic began in Wuhan.
Whether or not the Wuhan Institute of Virology’s coronavirus work had anything to do with causing the last pandemic — many virologists argue that a natural origin is more likely — there was widespread agreement among the experts I talked to that the huge collection of coronaviruses amassed before the pandemic had limited utility in developing treatments or vaccines once Covid-19 began spreading.
“After having done this work for 15 years, I think there’s little to show for it. As the intelligence community concluded, it’s plausible that it actually caused this pandemic, and to me that’s enough,” Weber told me. “We don’t have to be sure what caused this pandemic to reduce the risk of the next pandemic. It was of zero help in preventing this pandemic or even predicting this pandemic.”
“As best as I can tell, the only thing we needed for the vaccine was the prior work on the spike protein,” Esvelt told me, “and that did not result from any virus discovery or characterization in the lab.”
A USAID spokesperson disputed that claim. Predict, they told me, “advanced the current knowledge of several different viral families, including an understanding of where risks are and the human behavior leading to contact with animals that increases the potential for spillover. This information is being used by scientists to develop broadly protective vaccines and medicines, critical tools to have available for when/if a new coronavirus causes an outbreak in the future.”
Predict’s critics say that while almost any research can technically be said to have “advanced the current knowledge” of viruses, the benefits here are oversold to the public: nothing major, exciting, or especially promising came out of Predict’s viral discovery work — and the most valuable work Predict did was in testing humans near wildlife-human interfaces for diseases that had already crossed over into humans.
“Predict only discovered a single conclusive zoonotic virus that spilled over into humans — and this not through wildlife sampling, but from analyzing patient samples,” the recent Oxford and Georgetown paper on large-scale viral surveillance programs noted.
“Since the SARS-CoV-1 outbreak in 2003, numerous animal coronaviruses have been gathered and investigated, but this work did little to prevent the COVID-19 pandemic or inform vaccine design,” the researchers wrote, concluding the most valuable work was studies of MERS and SARS — the coronaviruses that had caused severe disease in humans.
The crucial zoonotic crossover work isn’t viral discovery
All of this isn’t to say that efforts to study zoonotic crossover aren’t hugely important, or don’t have a major role to play in pandemic preparedness. Much of Predict’s other work was hugely valuable — for example, research on reducing human-wildlife contact and enhancing international disease response partnerships.
There is no question that lots more preparedness work is needed to prevent the next pandemic. It’s just a matter of what work is best — and safest. “Ultimately, what makes one spillover event into a pandemic versus an isolated outbreak has a lot more to do with policies and health systems (i.e., community awareness, surveillance systems, rapid response capabilities) than it does about knowing ahead of time what sort of characteristics the virus has,” Georgetown biologist Claire Standley, one of the authors of the paper looking at surveillance programs, told me.
The paper ultimately highlights a more narrow approach as likely more cost-effective and lower-risk: focusing on response capabilities and human infections in areas where zoonotic crossover is a possibility. “Adopting such a highly focused approach for zoonotic risk prediction may not only reduce safety and security risks associated with the large-scale collection of wildlife viruses, but also generate more actionable insights — and likely at a lower price tag,” the paper concludes.
Esvelt’s ultimate takeaway? “Let’s not learn to make pandemics until we can reliably defend against them. Instead, we could take all of these funds that we were going to use to identify which particular viruses cause pandemics and pour it back into preventing spillover.”