How UVA’s Biocomplexity Institute Helped Map the Pandemic, From Jan. 2020 to Now

So, for the past 16 months and counting, it’s been all things COVID-19, all the time, for researchers at the institute’s Network Systems Science and Advanced Computing Division, which Marathe directs and where Lewis is a research associate professor.

Five to six researchers do the core work for the weekly forecasts the institute produces, Lewis explained, and another 20 provide “nuggets” to support its modeling; these include statisticians providing survey analysis and bioinformaticians analyzing the latest information on COVID-19 variants. One of the core researchers is a computational biologist, Przemyslaw (Przemek) Porebski, who had just started working with the institute when COVID-19 hit; his work has optimized the computations that provide data on variants and was “a key component of our team’s contribution to the CDC paper,” Lewis noted. (All 14 UVA researchers who contributed to the paper are listed at the end of this article.)

Marathe, UVA’s principal investigator for the Virginia Department of Health, said the Biocomplexity Institute was in touch with VDH and the state Department of Emergency Management early in 2020, and shortly thereafter started collaborating with the Department of Education.

The Art, Science and Value of Forecasting

UVA’s COVID-19 forecasting team has internal “sync” meetings, where researchers pitch ideas on what to analyze week to week, explained Lewis, who draws upon his own experience and epidemiological knowledge, as well as what’s trending in headlines, scientific literature, others’ modeling, intel from standing calls with agencies, and even social media. The latter may seem like an odd resource for scientists, but Chinese researchers were posting their findings in real-time on Twitter when COVID-19 first emerged, said Lewis, and “we began producing our weekly forecasts a few weeks later.”

“What has been unique to this pandemic is the degree to which rapid information sharing among the scientific community took place,” he said. The use of the “rxiv” sites – “archive” sites where scientists could pre-publish a paper awaiting peer review as a way to share information quickly in a fast-changing, emergency situation – “really helped everyone’s efforts,” Lewis said.

The team also worked with VDH from November to March on a contact-tracing paper. Marathe and Lewis advised the state to create smaller health centers that could be put up and taken down easily, instead of larger, more permanent testing and vaccination sites.

Recently, the team began collaborating on a new project related to COVID-19 variants. “We’re providing expertise and projections on how variants are evolving epidemiologically,” which could slow the progress in containing COVID-19, Marathe said. “We really hope we can control them.”

“The key part of this project is to how best use the genomic-sequencing capabilities VDH has to stay on top of the variants in the state so we can actively sense and interrupt their spread, rather than reacting to outbreaks that get discovered after they’ve been percolating for a while,” Lewis said.  

All along, the institute also has provided information to UVA’s senior leadership team to enable them to make decisions in the best interests of student health and safety. “We also provided projections to UVA Health and shared our models and support with other universities, including Virginia Commonwealth University, Washington and Lee, and James Madison University,” Lewis said.

“We are lucky to have this resource in the state,” State Epidemiologist Dr. Lilian Peake said. “I reached out to Madhav last spring to ask whether his team could provide a Virginia-specific COVID-19 model to help the commonwealth plan for potential scenarios. The Biocomplexity Institute had a track record of developing models for other infectious diseases, including some innovative approaches for predicting the timing of seasonal influenza.

“They quickly developed a model inputting Virginia data and have been exceptional partners. Our teams have worked together since that time, interacting multiple times each week. Madhav’s team has updated its techniques continuously to provide better information and has added analyses to provide insight on the effects preventive factors can have on transmission. Bryan expertly explains the output from these complex models on a weekly basis to hospitals and policymakers so they can understand the potential scenarios.

“This layer of analysis added on top of the descriptive epidemiology that VDH provides allows us to better monitor this emerging disease and develop contingency plans.”

How the Forecasts Are Made

The forecasts are created using a supercomputer that crunches a multitude of data, including national population demographics that drill down to the county level, disease surveillance, hospitalizations, patients’ ages, number of deaths, and more. “Using a supercomputer, it still takes hours for our projection runs,” Lewis said.

“Dry runs and forecasting with influenza allowed us to be ready to provide rapid models for COVID-19,” Lewis said. “Much like the military, we’re training to be prepared for a war we hope we never have to fight.

“We really were ready for this,” said Lewis, who said he considers Virginia a success story. The state has “managed this pandemic quite well. It helps that we have a governor who is a physician. Virginia’s high vaccination rate rapidity is to be commended. Health Commissioner Dr. Norman Oliver, Secretary of Health and Human Resources Dr. Daniel Carey, and Dr. Peake have done a good job.” And as a country, “We’re very lucky that we had vaccines in nine months,” he said.

The Way Forward – and Models for the Globe

In addition to getting vaccinated and continuing to practice non-pharmaceutical measures such as mask-wearing and hand-washing, Lewis recommends that we “build trust by talking to each other – it’s not a red- or blue-hat issue,” he said. “Seventy percent to 75% herd immunity doesn’t work if we’re segregated.

“COVID-19 will be around for the next couple of years. We won’t experience anything as bad as last year, but we’ll need booster vaccines, and we need to be a good neighbor to the rest of the world.”

The pandemic has provided a stark reminder that disease does not respect international boundaries. Lewis and his colleagues have set their sights on international containment efforts, and now provide data to CDC partners in Europe and India.

“We have adjusted our modeling and now contribute to a European CDC hub, providing weekly forecasting reports,” Lewis said. “We’re also watching India” and collaborating with its CDC academic partner, the Indian Institute of Science.

“We have a longstanding relationship with them,” Marathe noted. “We helped them build models similar to our own, and have written papers on how to open public transportation and how to distribute vaccines there.” Marathe hopes to soon explore similar collaborations in Africa.

The Biocomplexity Institute also advises the U.S. Department of Defense; typically, the military sets up distribution centers for testing and vaccinations domestically and abroad, and “we want to keep them safe,” Lewis said.

Team Science and Cultivating Disease Detectives

To be ready for public health crises, “You need people who understand policy and psychology, and that coordination is hard to do, especially in crisis,” said Lewis, whose own path to computational epidemiology began at Carnegie Mellon University, where he studied computational biology. But the “beauty of computation wasn’t enough,” he said; combining biology with computer science “clicked,” and his work in the Peace Corps in Africa after college led him to public health.

For Marathe, a native of India, it was an interest in science, specifically math and physics, and being “lucky to get to come to the U.S. and work with brilliant minds,” he said, that led him to his current pursuits.

“My mentors played a very important role in my life. They taught me humility, and that work is exciting and fun,” Marathe said. Now the director of the institute’s Network Systems Science and Advanced Computing Division, and a distinguished professor in biocomplexity and a professor of computer science, Marathe is a passionate advocate and practitioner of transdisciplinary team science, and is “very committed to teaching.” He has five Ph.D. students currently and has advised 25 Ph.D. and 24 M.S. thesis students to date in his 25-year career.

The institute also employs student researchers, who work closely with Marathe, Lewis and others on staff. One of those students – Joseph Outten, a 2020 UVA graduate and alumnus of the division’s innovative summer program called Computing for Global Challenges – is now a full-time employee of the institute and is among the 14 researchers who contributed to the CDC paper.

What’s Next?

“Yes, there will be future pandemics,” Lewis said. “There are so many more humans now. Urban areas are so dense.

“Another pandemic probably won’t happen for a while, but it will be sooner than the time span of the previous one in 1918.”

As was the case with COVID-19, the Biocomplexity Institute will be ready – not if it happens, but when, as Marathe likes to put it – and is leading two large NSF-funded projects, Expeditions and PREPARE. Both aim to find and stop disease outbreaks before they spread across the globe, to prevent another year like this one.

The 14 core Biocomplexity Institute researchers who contributed to the CDC paper are Akhil Sai Peddireddy, Pyrros A. Telionis, Anil Vullikanti, Jiangzhuo Chen, Benjamin Hurt, Brian D. Klahn, Bryan Lewis, James Schlitt, Joseph Outten, Lijing Wang, Madhav Marathe, Patrick Corbett, Przemyslaw Porebski and Srinivasan Venkatramanan.