On February 12, 2020, Professor Alessandro Vespignani gathered his team of research scientists in his 10th-floor office at Northeastern University to show them a potentially terrifying future. The half-dozen researchers huddled around a computer screen that listed the names of Asian and European countries, with a colored bar beside each name. Bar after bar, in dark red, surged to the right.
This was early in the story of COVID-19, then associated mostly with China; a mere 14 cases of the novel coronavirus had been identified in the United States, and the disease had gotten its official name just the day before. But Vespignani’s team at Northeastern’s Network Science Institute had just produced new simulations — the result of one million computer processors crossing known virus cases with airline traffic and other data. Each of the bars on Vespignani’s chart represented a surge in virus transmissions that was likely to take place over the weeks to come. If these projections were correct, the coronavirus would not overtake China alone. It would not be stopped elsewhere in East Asia. It would spread across the world. “If this is what it looks like, this is a pandemic, and this is going to be a really serious one,’” Vespignani recalls telling his team.
Five months later, I met Vespignani in his otherwise empty computing lab; the network science team had been working largely remotely since mid-March. A large bottle of Purell sat on his desk, atop many archaeological layers of papers. Seated on a semi-circular couch by the lab’s windows, he recounted the gravity of that mid-winter moment. “That was really one of the worst days of my life,” he told me.
Yet back in February, the graduate students and postdocs in his lab never got a sense of panic or despair from the wiry professor with close-cropped black hair. “I think he probably used the words, ‘Oh, wow,’” says Jessica Davis, the Ph.D. student who had sent Vespignani those doomsaying graphs. “He was calm, relaxed, but then pushing on to the next thing.” At the time, Vespignani told his team to run more simulations. (“We didn’t want to create alarm without being sure,” he recalls.) Six days later, Davis sent him another email, with the subject heading, “PROOF.” Vespignani warned the Centers for Disease Control and Prevention.
The modeling Vespignani and his team did in early 2020 helped drive an international effort to predict the pandemic’s course. Mathematical modeling of epidemics is more than a century old, but it made vast strides during the 2010s, thanks to advanced new computational models that Vespignani played a part in developing. Using an extraordinary amount of computing power, his team takes in data from such sources as airline flights, urban traffic patterns, and cellphone locations, crosses it with population data and maps, then crosses both with a disease’s known characteristics to produce scenarios and predictions of an epidemic’s spread. Vespignani’s Laboratory for the Modeling of Biological and Socio-Technical Systems — known as the MOBS Lab — is among dozens of epidemic modeling teams across the world. But MOBS’ Global Epidemic and Mobility model is more ambitious than most: it seeks to simulate the behavior of every single one of the world’s seven billion people.
Since winter, Vespignani has been sharing his lab’s results with the World Health Organization, the Gates Foundation, the White House, and the CDC — which weaves his work into a weekly ensemble forecast, now based on 30 modelers’ individual projections. As his team continues its super-accelerated, rapid-response science, Vespignani hopes that ad hoc efforts like his will lead to permanent national and international efforts against infectious disease. Globalization is making pandemics more likely, he warns, and the world needs to react even faster next time.
“This is really a war,” Vespignani told me in his office that midsummer day. He was dressed as if for a casual Friday, in a small-collared sport coat, an untucked but sharply-cut black button-up shirt, dark blue jeans, and Nikes with a fluorescent green swoosh. A couple of days’ worth of gray stubble showed between the straps of his light-blue paper mask. He took off his glasses. Behind the mask, he looked tired. “I never thought I’d see what I’m seeing. For anything so unprecedented, you have to go back to World War II.”
Forecasting pandemics is not at all what Vespignani set out to do as a young scientist, who got his Ph.D. in theoretical physics at the University of Rome in 1994. Vespignani wrote his dissertation about fractals — the complex geometrical shapes that nature forms, which can be seen in sandpiles and tree branches. Benoit Mandelbrot, known as the “father of fractals,” recruited Vespignani for a post-doctoral job at Yale.
“That was the first time I went out of my country,” Vespignani says. “I liked the academic U.S. system, the fact that you can be entrepreneurial, you can be very young, and if you have ideas, you find support.”
Back in Italy, in a research job in theoretical physics, Vespignani grew fascinated with the work of Albert-László Barabási — now a Northeastern colleague — who was mapping the internet. “He was showing those incredible, complex structures of the internet, and I really got fascinated by it,” says Vespignani. Starting in 2001, he and a collaborator published a series of academic papers on how computer viruses spread on complex networks — the first appearances of the word “epidemic” in his curriculum vitae.
Vespignani landed a tenured job at France’s National Center for Scientific Research, but his interests continued to draw him to digital epidemics. “I started to grow a little bit uneasy, because I was working mostly in computer science in a physics department,” he says. Then a former University of Rome classmate working at Indiana University convinced Vespignani to join him there, to help create a school for computational and information technology. “My wife looked at me like I was crazy. She said, ‘Why, from Paris, should we go to Indiana?’” Vespignani recalls with a laugh. “I said, ‘This is, for me, a chance to move to the next level in terms of research.’”
“He said, ‘It’s going to be bad; things are going to get shut down.’ He laid out remarkably well what would happen in the next two months.”
David Lazer, political science and computer science professor and Vespignani‘s colleague at Northeastern’s Network Science Institute
In Indiana, Vespignani turned his attention from computer networks to transportation networks — and from computer viruses to actual viruses. When the H1N1 flu pandemic broke out in 2009, Vespignani and his team put their algorithms to work, accurately predicting, months in advance, when H1N1 would peak in 42 countries.
After seven years at Indiana, Vespignani got another enticing offer: the chance to work with Barabási at Northeastern and direct the nation’s first Ph.D. program in network science. “They called me and [asked], ‘What are your ideas?’” Vespignani recalls. He proposed an interdisciplinary institute that applied network science to physics and mathematics, but also political science and economics.
“‘This is my dream,’” he says he told his future colleagues. “‘I want to do something that is really right, and if it doesn’t work, it doesn’t work.’ … The response here was terrific.”
Today, Vespignani’s official title reflects his interdisciplinary role — he is Northeastern’s Sternberg Family Distinguished Professor of Physics, Computer Science, and Health Sciences — and the Network Science Institute takes up three floors of an office tower in the Christian Science Plaza in Boston’s Back Bay neighborhood. Vespignani’s lab has no beakers, lasers, or semiconductor instruments; the team of 15 that he supervises, mostly data scientists, work at cubicles that look a lot like most modern professional offices.
On the mid-July day when I met him, Vespignani’s corner office looked frozen in time. Math formulas were scrawled on his window glass, with the red brick townhomes of Boston’s Back Bay and South End below and beyond. On his desk sat journal articles the lab had published on the virus, along with one of the last sets of bar charts his modeling team had produced in early March — a continuation of the work they had done on that first fateful day. It showed seven countries, including the United States, with a near-100% chance of reaching 10,000 COVID-19 cases by March 31. These bars were blue; as the epidemic grew, Vespignani told his researchers to stop using red in their charts. “At a certain point, I told the team, ‘Guys, we need to change color, because these are too terrifying,’” he recalls. “The message is the same, but I don’t want to have something that looks like the end of the world.”
COVID-19 wasn’t the first time that Vespignani’s team has taken on a virus mid-epidemic. When the Ebola virus broke out in West Africa in 2014, the office shifted to a phase Vespignani refers to as “tactical” mode, when science doesn’t proceed at its usual cautious, peer-reviewed pace.
The lab worked to produce intelligence estimates about the viral enemy in real time, using an “agent-based” modeling approach that Vespignani helped to develop. It’s a different approach from the “compartmental” modeling that most of his peers at other institutions use. Compartmental modeling uses equations to calculate trends, such as the average transmission rate of a virus, within a large population, such as a state’s or a city’s. Vespignani’s team, by contrast, tries to simulate the behavior of individual people — sometimes, all seven billion of them.
Agent-based models aren’t necessarily more accurate, says Ruth Etzioni, a biostatistician at the Fred Hutchinson Cancer Research Center in Seattle, but they offer an advantage over compartmental models. Their massive data processing and fine-grained detail allow them to estimate the effects of a wider variety of interventions than a compartmental model can — to project what kinds of actions, from governments and individuals, might best stop a virus’s spread. “An agent-based model is like SimCity, trying to do a virtual version of the population,” she says. “In principle, it can capture things like clusters of cases.”
In the case of Ebola, agent-based modeling allowed Vespignani’s team to play with different possibilities: What would happen if public health officials did more vigorous contact tracing? What if they built more hospitals? Vespignani says his projections helped international leaders make difficult decisions about how much to isolate the West African countries where Ebola was spreading, helping them stop the virus at international borders without wrecking stricken nations’ economies — or blocking aid from coming in.
The mosquito-borne Zika epidemic, which broke out in 2015, posed an even bigger challenge. In this case, Vespignani and his team didn’t just have to model seven billion humans. They also had to predict the behavior of many more billions of mosquitoes, along with their reproductive cycles.
“I discovered that through machine learning, you can have a very detailed, accurate map of mosquito presence and abundance,” Vespignani says. U.S. public health officials used Northeastern’s Zika projections, which forecasted virus hot spots, to single out locations for vaccine trials.
Since then, Vespignani’s team has helped to forecast annual flu seasons for the CDC and analyzed an Ebola resurgence for the WHO.
He co-wrote a book called Charting the Next Pandemic, published in early 2019. A massive world map included in the book still hangs on his lab’s wall. It charts a theoretical influenza outbreak: grey branches fanning out from Johannesburg, South Africa, the origin of the imagined disease. Blue dots of infection bloom worldwide, clustered in dense population centers with international connections: cities in India, China, Europe, the United States.
It was early January 2020 when Vespignani’s team was first drawn into surveillance of the new respiratory illness spreading from Wuhan, China. The WHO emailed epidemic modelers on several continents and gathered them on a call that, in Boston, took place at 5 a.m. Vespignani’s team jumped to work, reporting its preliminary projections of the virus’ spread on an infectious-disease research website in mid-January, warning of “likely a large number of undetected/untested cases.”
In late January, during an evening meeting in his office, Vespignani asked Davis, the Ph.D. student, if she’d ever seen the movie Contagion. She hadn’t. The 2011 Steven Soderbergh film imagines an airborne virus outbreak that becomes a pandemic. Vespignani told his researchers to watch Contagion to “prepare.” At first, Davis thought Vespignani was joking, but he wasn’t. Though the film imagines a virus much deadlier than COVID-19, the screenwriter had consulted several epidemiologists while writing it. Vespignani told his team that it contained insights about how the government and the public might respond to a pandemic.
Vespignani and his colleagues released a preliminary paper in February projecting that travel restrictions to COVID-19–affected areas would have “modest effects,” but that “transmission reduction interventions” — public-health efforts such as quarantines and self-isolation — would do more to mitigate the epidemic. The paper, formally published in the journal Science in April, had been cited in 697 other papers as of late August. Jeffrey Shaman, a Columbia University environmental health sciences professor who has also released key studies of the COVID-19 virus, says the Vespignani team’s paper provided early insight into how the virus spreads. “It’s something that was incredibly important to see and use at the time,” Shaman says.
That month, David Lazer, a political science and computer science professor at the Network Science Institute, often noticed Vespignani pacing. “I was staying late at work one night in mid-to-late February, and Alex was there, standing at his office door a moment,” Lazer recalls. He asked Vespignani about the epidemic. “He said, ‘It’s going to be bad; things are going to get shut down; the numbers in New York are going to get very, very bad very, very quickly.’ He laid out remarkably well what would happen in the next two months.” Lazer quickly stocked up on frozen food, pasta sauce, and toilet paper. He also switched the focus of his work from studying political misinformation to studying COVID-19 misinformation.
In the Boston area, Vespignani became an early-warning sentinel. In February, he advised Northeastern officials that they’d likely have to carry out a shut-down. In early March, Joe Curtatone, the mayor of Somerville, Massachusetts, attended a meeting at Northeastern about Vespignani’s modeling. What he heard was “much more than was being described publicly in the media,” Curtatone recalls. He walked away with an increased sense of urgency: “Days mattered, if we were going to avoid the collapse of the health care system.” Curtatone asked Vespignani and his colleagues to meet with a group of mayors and town administrators. After the meeting, the mayors signed a letter urging Massachusetts Governor Charlie Baker to close all schools in the state — which he did three days later.
“Without that help, that guidance, we would have acted later,” says Curtatone. “We wouldn’t have done enough soon enough. I submit we wouldn’t have had as great a success. More people would have died.”
Since then, Vespignani’s team has continued its regular surveillance of the virus. Its modeling of the pandemic’s so-called “cryptic phase” — when the virus spread mostly undetected — was a major source for a New York Times report in April and its mapping project in June. Every week, his team’s projections contribute to the CDC’s ensemble forecast, which Vespignani likens to the National Hurricane Center’s cone-shaped projections of a hurricane’s possible paths. “There are all of those little spaghetti that define the cones,” he says. “Each one of those spaghettis is a model, and you compound all those models together to get an uncertainty, and that cone is the cone of uncertainty.”
The modeling work has required massive computing power, the equivalent of a million computer processors working at once. Since January, Google Cloud has allowed Vespignani’s team to run simulations on its server networks — as the company did when Northeastern was modeling the Zika epidemic. Grants from a Northeastern University trustee, Jim Pallotta, and a trustee emeritus, Sy Sternberg, allowed the team to continue its work on the pandemic by freeing some of the staff from restrictions in grants that normally fund their work.
Vespignani believes computer modeling has made a difference in fighting the pandemic; network scientists’ early warning calls gave decision-makers more time to enact countermeasures, such as social-distancing regulations and business closings. If a virus like COVID-19 had struck 15 years ago, before today’s forecasting technology, the national numbers could have been far worse, he says. “Every time you are delayed by one week,” Vespignani says, “that means several thousand deaths more.”
In the future, Vespignani argues for permanent national and international early-warning systems, the public-health equivalent of the National Hurricane Center. It will take a global effort, he says, to “monitor viruses constantly, possibly in species that we are not often in contact [with],” to “at least see signals and warnings,” and to develop vaccines more rapidly.
He’s looking forward to the day when a vaccine, or another way to cope with COVID-19, frees humanity to resume living as we did before the pandemic. “I miss the personal contacts,” Vespignani says. “I’m Italian — I want to shake hands. I want to hug people. I want to go in a restaurant and talk with people without having the fear that my mask is not properly placed.”
And Vespignani misses travel, especially to conferences. A dozen lanyards hang from a peg in his office, from international conferences he’s attended on Ebola vaccines and epidemic modeling. “We can achieve a lot by video conferencing, but it’s not the same. From time to time, especially in my job, you need to sit in front of a whiteboard together.”
It’s not lost on Vespignani that the ease of global travel and communication, which brought him from Italy to the U.S. and made his career possible, also makes humanity more vulnerable to pandemics. “Now something that happens in China is coming here in one month or less,” he says. “We are all one flight away.”
But he hopes this pandemic is preparing the next generation of scientists — including the young network scientists in his lab — to fight future outbreaks with speed, clarity, and data. “There will be a next time,” he says. “I hope that they will be ready.”
Published on
