The 2014 Ebola outbreak in West Africa killed more than 11,000 people and was the deadliest outbreak since the discovery of the virus in 1976.
A New England Journal of Medicine study linked the start of the outbreak to a two-year-old in Meliandou, a remote village in Guinea, a country that had never before seen a case of Ebola. The type of virus was identified as the deadliest strain, found in Zaire—a country thousands of miles away.
How did it get there? The likeliest answer: bats.
A growing number of researchers are exploring the ecological dimensions that contribute to outbreaks of zoonotic diseases like Ebola that spill over from animal carriers to human populations. According to a report published by the Pulitzer Center, zoonotic diseases account for 60 percent of the roughly 400 emerging infectious diseases that have been identified since 1940. There is evidence that human activities such as deforestation are impacting animal migratory patterns as the search for resources is bringing animals into closer contact with humans.
Javier Buceta, associate professor of bioengineering, Paolo Bocchini, assistant professor of civil and environmental engineering, and postdoctoral student Graziano Fiorillo have created a modeling framework that takes a zoonotic perspective on Ebola. The model considers the ecological dimensions that drive bat migration patterns and could help predict a bat-transmitted outbreak of Ebola among humans.
“Our results highlight the importance of coupling information about seasonal resources with infection dynamics to predict when humans might be most at risk of contracting the virus through contact with bats,” says Buceta. “In our model, the appearance of outbreaks is tightly linked to fluctuations in environmental conditions which have an impact on both bat migration patterns and infection rates.”
The team’s approach works by tracking the migratory patterns of bats, which are believed to be a main carrier of the Ebola virus. Bats, in this instance, are the reservoirs of Ebola. This means that they are carriers and transmitters of the virus, though it does not cause them harm.
Buceta worked with a Lehigh undergraduate student, Kaylynn Johnson (currently at the University of Pennsylvania School of Veterinary Medicine), and developed a mathematical model to understand how Ebola infection dynamics in bats are linked to environmental factors. The results of that research were published in the journal PLoS in 2017. Later, Buceta and Bocchini combined that approach with satellite information and parameter sampling techniques. Their framework integrates data and modeling to predict the conditions linking bats’ behavior with the outbreak of Ebola. They have detailed their work in a paper titled “A Predictive Spatial Distribution Framework for Filovirus-Infected Bats.”
The model utilizes information on bat birth and death rates, the rate of infection of bats with the Ebola virus and recovery rates, bat mobility, seasonal changes and information about the availability of food and shelter to forecast bat infection peaks in a given region.
An analysis of the data from the region near Meliandou—where the 2014 outbreak began—revealed two yearly peaks of infection at Meliandou coinciding with the birthing seasons. Indeed, when the researchers retroactively analyzed the data from that time and location, their model predicted a peak of Ebola-infected bats in Meliandou during the months when the outbreak began.
By contrast, when the researchers worked with similar data from Bamako, Mali—a region nearly 400 miles away from Meliandou and with very different climate conditions—their approach forecasted only one peak.
“Such findings strongly suggest that environmental factors play a key role in the spread of the Ebola virus among bats,” says Buceta.
This predictive tool could be used to deliver guidance about the specific locations and periods of the year during which an outbreak is more likely to appear due to bats. It could also help reduce the risk of future spillovers from animals to humans.
The spread of Ebola infection from bats to humans is more likely to happen when the density of infected bats reaches a certain level.
“The availability of food and shelter plays a vital role,” says Buceta. “Our analyses reveal that the more resources that are available, the lower the infection rate—and vice versa.”
According to Buceta, when resources are low, more bats gather in the same place to compete for the resources they need to survive.
Establishing how to measure the key environmental factors driving resource-related bat migration was key to developing the model.
To do this, the team used environmental and climate data retrieved using the Google Earth Engine tool to access databases from the Land Processes Distributed Active Archive Center (LP DAAC), one of several discipline-specific data centers within the NASA Earth Observing System Data and Information System (EOSDIS).
Bocchini, a civil engineer, had been working with smart sampling techniques to resolve parameter fluctuations pertaining to his research on structural engineering and regional hazards. Through that work, he developed a highly efficient computational technique that addresses probabilistic big-data problems and enables researchers to analyze a small subset of truly representative cases.
“We needed to study the random fluctuations of available resources over the entire African continent at high resolution; it was a massive computational and probabilistic challenge,” says Bocchini. “We recognized that from a mathematical point of view, the problem is similar to the random propagation of seismic waves in a region subject to earthquakes, and we could adapt our tools.”
They applied Bocchini’s sampling technique to resolve the uncertainties in the data and to establish useful parameters for measuring resource availability, given fluctuating conditions over time and geography.
After establishing the parameters, they were able to input data regarding temperature, humidity and other factors
“We could then predict the concentration of infected bats one might expect to find given those particular conditions,” says Buceta.
For their next steps, the researchers plan to apply the “full force” of Bocchini’s novel sampling technique, called Functional Quantization, and conduct extensive numerical simulations on a High Performance Computing platform. They will also incorporate socioeconomic, cultural and demographic information to understand how such factors impact Ebola transmission from bats in a given region. They intend to create a tool called Predictive Analysis of the Risk of Ebola Outbreaks (PAREO) that will assist in predicting bat infection dynamics and Ebola outbreaks.
“As a civil engineer,” says Bocchini, “I look forward to closing the loop and studying how our society, our urban fabric and our infrastructure systems play a role in disease spreading in developing and developed countries.”
The project will be a multi-disciplinary collaboration that includes Shin-Yi Chou, a colleague from Lehigh’s College of Business and Economics who specializes in health and health care economics. It is also one of the first success stories of the university-wide Probabilistic Modeling Group, of which Buceta and Bocchini are founding members. The group facilitates interdisciplinary collaborations like this one.
The team’s method demonstrates the importance of examining the big picture when trying to find answers to complex questions.
“It’s clear that humans have our own role in the process,” says Buceta. “Through our actions, we not only disrupt sensitive ecological systems, but also increase our own chance of becoming infected with serious—sometimes deadly—diseases.”