Challenges often shake things up and inspire people to look at a problem from another angle. Just ask University of Arizona professors Joceline Lega and Heidi Brown.
Their prize-winning work to predict the spread of the chikungunya virus (CHIKV) in the Americas could have implications that extend to other diseases, including Zika.
Lega, a mathematics professor, and Brown, an assistant professor in epidemiology and biostatistics, won $150,000 in the Defense Advanced Research Projects Agency (DARPA) CHIKV Challenge last year, developing a mathematical model to forecast the infection growth rate as the disease emerges in new countries.
Global health organizations provide surveillance of infectious disease outbreaks but struggle to predict with enough certainty the spread of a disease.
That truth is what prompted DARPA to hold the CHIKV Challenge.
The Defense Department itself is responsible for monitoring the health of military personnel and others around the world. If military and public health authorities could move beyond basic surveillance to accurately predict where and when a disease will appear, they could deploy supplies and launch educational campaigns well in advance.
Best-case scenario: They could get out in front of it and stop it in its tracks.
“We are on the cusp of enabling a revolutionary improvement in disease forecasting, in much the way that weather reports transitioned from surveillance to forecasting,” said Col. Matt Hepburn, DARPA program manager for the CHIKV Challenge.
A Direct Approach
In the case of an infectious disease, modeling experts must consider factors such as vectors, hosts, how the disease develops in an individual, and how it is transmitted between people.
The DARPA CHIKV Challenge sought methods to forecast outbreaks and the potential spread of CHIKV throughout the Americas.
But the CHIKV Challenge demanded something specific: Extrapolate the number of reported cases in each country. This allowed competitors to try something new.
“The data was telling us that one could predict average trends in a much simpler fashion,” Lega said. “I don’t think we would have followed this route from the start, had the challenge not put emphasis on developing a solution that addresses a specific question.”
The challenge jolted the disease forecasting community, including the Arizona team whose model was the most accurate in predicting the number of reported cases CHIKV over the course of six months.
More than that, it gave them a head start in efforts to apply what they learned to the Zika virus.
Here We Go Again
In addition to energizing citizen scientists, a good challenge has staying power and produces solutions that can be adjusted and applied again and again.
Just as DARPA announced the winners of the CHIKV Challenge, the Pan American Health Organization (PAHO) issued an alert for the first confirmed case of Zika virus in Brazil.
Since then, the rate of infection has increased, and the CDC announced that enough evidence has accumulated to conclude that Zika virus infection during pregnancy is a cause of microcephaly and other severe fetal brain defects.
The Zika virus is transmitted by Aedes mosquitoes, the same kind that carry CHIKV. The PAHO reports that 42 countries and territories have confirmed local, vector-borne transmission of Zika virus disease in the Americas, including the United States, where authorities have now detected locally transmitted cases of Zika virus.
The biggest obstacle to forecasting so far is a lack of data. PAHO was posting weekly counts of CHIKV cases, which helped to inform the University of Arizona researchers’ model.
“We suspect that the model should work equally well for forecasting the number of cases, the duration of the outbreak, and when the outbreak peaks for Zika as it did for chikungunya,” Brown said.
She and Lega are studying the differences between CHIKV and Zika with the piecemeal information available in an effort to better understand the Zika outbreak and any that come after it.
Thanks to the CHIKV Challenge, they have more resources to aid them in their research.
The Winning Duo
A challenge competition has a way of opening doors to new opportunities and joining people from different backgrounds to work towards a common goal. The CHIKV Challenge was the impetus for the winning partnership between Lega and Brown. They had worked together before, but the CHIKV Challenge allowed them to explore a new opportunity and led to a string of new collaborations that didn’t exist before the competition.
From left to right: DARPA Program Manager Col. Matt Hepburn, winners Joceline Lega and Heidi Brown, and DARPA Director Arati Prabhakar.
“As an epidemiologist working with a mathematician, this challenge has led me into a more quantitative direction,” Brown said. “Working with Joceline has helped me to see my own research area from different perspectives.”
As for Lega, she just finished spending part of a sabbatical in Brown’s department to learn more about epidemiology and public health.
“This is a new research direction for me, and I hope this is just the beginning of a very fruitful collaboration,” Lega said.
The two now are expanding their research team to include other faculty and students at the University of Arizona.
But the real team created by DARPA’s challenge goes well beyond the work at any one university.
In addition to Lega and Brown, 10 other teams won cash prizes for their models in the CHIKV Challenge. Participants spanned a range of disciplines and included specialists in public health and infectious disease, as well as mathematics, ecology, computer science and bioinformatics.
None of them had worked with DARPA before, which highlights one of the agency’s goals in launching the CHIKV Challenge — to tap the knowledge-base of thinkers previously unknown to the agency.
“This forward-thinking collaboration is exactly what it will take to stay ahead of the global threat that emerging diseases pose,” Hepburn said.
In May 2015, the winning teams all came together with DARPA and officials from other government agencies and public health organizations. Representatives from Los Alamos National Laboratory (LANL), who helped evaluate challenge submissions, also attended.
“It was a fruitful experience to see a room full of academics interacting with representatives from the Defense Department, and the government, more broadly on a very real health problem,” Brown said.
The assembled crowd talked about the competition, what they learned from it and what their now-expanded community could do next to push the boundaries of infectious disease forecasting.
Prize competition experts like to say that the end of a challenge is really just the beginning. The real work begins after winners are announced.
This certainly applies to the CHIKV Challenge, especially as another mosquito-borne disease is on the rise.
DARPA has continued to collaborate with winning teams since the conclusion of the competition as the new partners work to update CHIKV and Zika projections.
The challenge also has led to an ongoing dialogue between DARPA and public health experts on how forecasting infectious disease can improve public health decision-making in the Americas, Hepburn said.
“The challenge showed that models are more accurate when they incorporate key data streams, such as meteorological, geographic, human, and vector parameters,” he said.
With that in mind, DARPA has worked with LANL and the top six winners of the challenge to write a report looking at existing methods, technological gaps and other critical factors for making successful predictions. The report has been submitted for publication in an academic journal.
In another effort to share knowledge about the problem and potential solutions, DARPA has addressed how case reporting stats are compiled and communicated in an effort to improve modelers’ access to data.
The team from the University of Arizona also continues to adjust its model, share it with others and make it as useful as possible for different scenarios and changing conditions.
Lega has been working on automating the way the model estimates numerical inputs, to make it easier for others to use. She also designed a more theoretical version of the model using information from previous flu seasons and current flu data. The latter effort will help determine how much the CHIKV approach can be extended to other infectious diseases.
There are still many unknowns, and the answers won’t come easy, but the knowledge gained through the experience of the CHIKV Challenge has put researchers on solid footing as they move forward.
Like any tough problem, there probably won’t be a silver bullet.
“I don’t believe there will be one model,” Brown said. “Rather, there will need to be a suite of models, each addressing a piece of the puzzle. As with any good research question, this challenge has generated many more avenues for research.”
And the prize money certainly helps explore those new avenues, Lega said, “including more unconventional ways of thinking, which would not normally be funded by standard grants.”