Well before COVID-19 struck New Mexico, New Mexico was striking COVID-19. Los Alamos National Laboratory (LANL) began its research on COVID-19 in late January, one of several national labs to position itself as an early mover in fighting the pandemic research with high-performance computing. As the pandemic nears the one-year mark, LANL has published a retrospective highlighting the wide variety of COVID-19 computing projects that the lab hosted over the last 11 months.
Pivoting LANL epidemiology to COVID-19
Some LANL researchers, like scientist Dave Osthus, had an easy pivot to COVID-19 thanks to ongoing virus research that preceded the pandemic. Osthus and his colleagues provide weekly infection forecasts during flu season using a model so successful that it had won awards in previous flu seasons. In the midst of submitting this model for this year’s CDC FluSight flu modeling competition, the competition was canceled as the pandemic began to sweep the world.
The team iterated on their model to accommodate COVID-19 forecasting, eventually settling on a minimalist model that ingests infection data, death data and population sizes to continuously forecast new infections. In contrast to the flu models, however, the COVID-19 model uses LANL’s Darwin supercomputer. “As of now,” Osthus said, “our system is one of only a handful of models being used by the CDC that has consistently outperformed the baseline models.”
Similarly, LANL computational scientist Timothy Germann had been developing EpiCast – an agent-based virus model – for around 15 years before the pandemic struck. EpiCast simulates a virtual city populated by individuals in different demographics with different professions, then studies how those individuals behave and how that behavior spreads a virus with given parameters. Originally built to study smallpox and the avian flu, Germann and his colleagues have now adapted the tool to study how COVID-19 would spread through New Mexico. EpiCast has been used to guide public policy decisions in the state.
LANL information scientist Ashlynn Daughton also pivoted her work, which had previously focused on analyzing travel cancelation indicators in social media posts to understand public sentiment on the Zika virus. Building on this foundation, Daughton created an algorithm that pulled social media posts from Twitter and reddit that analyzed sentiment around social distancing, hand-washing and mask-wearing. “In the long term, it will be interesting to see whether living in a place where online misinformation about COVID-19 is widespread makes you more likely to buy into that misinformation and put yourself at risk,” Daughton said.
“With our history in epidemiological modeling and research, it was a natural pivot for us,” said associate lab director J. Patrick Fitch, who leads the lab’s coronavirus response. “From an impact point of view, the modeling we’re doing with computers has changed our country’s understanding of this pandemic.”
Understanding the biology of the virus
Others started from scratch, trying to understand the mysterious origins and behavior of the virus. LANL mathematical biologist Ruy Ribeiro worked with his colleagues to model how SARS-CoV-2 affected human organs, primarily how it migrated from the upper respiratory tract to the lungs. Using their model, the team identified when people proved most infectious and identified when in the course of an infection certain therapeutics could prove more (or less) effective.
Theoretical biologist Bette Korber, meanwhile, examined the various variations of the virus using geographic information from GISAID virus genome samples. She and her colleagues concluded that a specific variation (“D614G”) had become the most prevalent variation.
Similarly, computational biologist Thomas Leitner developed a “genetic tree” for COVID-19, tracing its genetic history back to its ostensible origins in China using the same GISAID database. “There are fewer atoms in the universe than the number of topologies I can create with 49,000 genetic variations of this virus.” Leitner said. “I’ve definitely got my work cut out for me.”
Bridging the gap for hospitals
Yet other LANL researchers focused on ensuring hospitals were well-equipped to handle COVID-19 patients. LANL data analyst Paolo Patelli and computational scientist Nidhi Parikh, for instance, created a predictive tool to anticipate demand for ventilators in a given state. “Since then,” Parikh said, “we’ve expanded the program to include personal protective equipment, medications, and almost anything a hospital needs to help COVID-19 patients.”
… and more.
To read about even more ways that LANL fought back against COVID-19 this year, read the article by LANL’s J. Weston Phippen here.