A particle physicist ventures into epidemiology
Edward Wenger was at CERN in 2009 for the Large Hadron Collider's first proton–proton collisions. By 2012, when the Higgs boson was discovered there, he was at the Institute for Disease Modeling (IDM). How did he go from being a particle physicist in Switzerland to leading a multidisciplinary team in Seattle that aims to eradicate malaria?
Wenger earned his PhD at MIT, working on the PHOBOS detector based at Brookhaven National Laboratory's Relativistic Heavy Ion Collider. When he finished, he went to CERN, just in time for the heady days of devastating mishaps and jubilant success at the LHC. And then he took stock.
Yes, he loved physics. But he saw that if he stayed in academia, he wouldn't be able to choose where he lived. "This is part of what a lot of people in physics jokingly refer to as the coupled oscillator problem," he says. "My wife and I are both from Seattle, and we wanted to raise our family back here."
Edward Wenger (center) with Samwel Onditi (left) of PATH in Kenya visit Calvester Machila, district malaria officer in Sinazongwe, southern Zambia, in October 2012, to observe mass screen-and-treat rounds. CREDIT: Elyse Callahan
So he returned to his hometown and started looking for a job. "It turned out that this was a great opportunity for me to make a conscious choice about what I wanted to do," he says. "I wanted to use the skill set I had coming out of my physics career. And I wanted to make an impact on the world and to pursue interesting and challenging problems."
The IDM started out as part of the Intellectual Ventures Lab, the R&D branch of a company started by Nathan Myhrvold and Edward Jung, both formerly of Microsoft. Intellectual Ventures holds among the most patents of any private company. The lab focuses on areas that have either commercial prospects or humanitarian goals.
According to the IDM website, the institute "develops detailed, geographically specific, and mechanistic stochastic simulations of disease transmission" and aims to "determine the combination of health policies and intervention strategies that can lead to disease eradication." The IDM spun off in February from the lab, but remains under the parent company.
Other work at the Intellectual Ventures Lab, which was founded in 2007 and now has about 100 employees, includes disease diagnostics; mosquito breeding to study reproduction, disease transmission, and the selective killing of females; methods for safe milk transportation in rural areas of the developing world; and an instrument shop for prototyping inventions.
Wenger spoke by phone with Physics Today about his new line of work.
PT: How did you happen to go to the Institute for Disease Modeling?
WENGER: It turns out there are a whole bunch of really exciting opportunities—in software, in aviation, in banking. It was pretty much by chance that I heard about the Intellectual Ventures Laboratory. I interviewed here, and it was a really appealing position because on the one hand I got to make use of my existing skill sets in quantitative data analysis and with the software. And on the other hand I get to do something that has the potential to have a huge impact on people's lives. In a way, that's the missing piece in physics.
When I was a physicist, people loved the LHC. And when they found out that I was based in Geneva, they'd say, "That's so cool," and want to have these long conversations about it. Eventually, almost everyone would ask, "So, how does this affect me?" And I'd say, "It doesn't, really. But it's really exciting research. We now know what matter was made out of a microsecond after the Big Bang better than anyone ever did before."
Now, doing disease modeling, we really have a huge impact on global health decisions. There is a lot of room for rational allocations of resources that actually save lives.
PT: Describe disease modeling and what your work consists of.
ENGER: Generally speaking, disease modeling is trying to capture the relevant dynamics of disease transmission through a set of equations that we write up into a piece of software. This typically entails researching the relevant scientific literature and designing software architecture that addresses specific questions.
Across all of our diseases we include demographics and migration. For malaria, we built up a mosquito model. That includes larval habitat dynamics, how that depends on the weather, and on mosquito feeding cycles, and on the various decision points for mosquitos—whether they feed on animals or on people, whether they feed indoors or outdoors. How the acquisition of human immunity impacts both the burgeoning of disease and the detectability in individuals and how likely they are to seek care all come into play. These are important factors in the impact of intervention. Having built up that information, the whole idea is to apply a model to very specific questions.
PT: What questions, and who is asking them?
WENGER: There are a whole bunch of stakeholders in global health. Some of the questions come from the [Bill and Melinda] Gates Foundation; there are country-level ministries of health that ask us questions; and global health programs, such as PATH. On our malaria work, we are actively partnering with groups doing work in Zambia, Kenya, Mozambique, the Solomon Islands, Senegal, and parts of Southeast Asia. [The IDM's] polio, HIV, and tuberculosis work is deeply involved with research groups in Nigeria, South Africa, and China.
[There are] a couple of broad categories of questions we get: What is the optimum allocation of resources? Is it bed nets or insecticide spraying? We also ask questions like, What existing tools are necessary and sufficient to achieve elimination? How does that depend on the specifics of the larval habitats and the mosquito behavior in a specific setting? What are the impacts of future technologies? There is a lot of work going on to develop better diagnostic tools.
Having built up a body of software, we initialize it with a set of parameters that we think are characteristic of a specific site. We may run a baseline scenario keeping track of how many people are infected through time, and how many severe cases and deaths we've seen as a function of time. Then we can sweep over different interventions, and say "What if we gave out bed nets to 20% of the people? 40%? 80%? What if the average person gets 1 infectious bite a year? 10? 100? We sweep over that type of space by submitting tens of thousands of jobs to supercomputers. By analyzing the output in aggregate we start to get answers to the questions.
What we are really aiming to do is to get a fast feedback cycle between surveillance data that is collected as input to our model and the output that is coming out of that [model]—to let groups make more rational allocation of resources or take better next steps forward.
PT: Do current world events play a role? For example, the recent outbreak of polio in Syria?
WENGER: The [polio eradication] endgame considerations certainly are flavored by current events when you try and understand how effectively can we mop up outbreak events.
PT: Have you seen cases where the predictions from your models have had a real impact?
WENGER: Yes. In our ongoing work in southern Zambia. We've been working with them to analyze their surveillance data coming out of their mass screen-and-treat activities. [The data include] which individuals they've reached, and what the geographical patterns of infections are, the age patterns of bed-net ownership and usage, and so on. [Our partners there] are using our results to decide on operational changes—whether there's a different mode of distribution, or whether to switch to different drugs that have longer prophylactic effects.
PT: How do you avoid people in Zambia or other countries feeling that your work is an imposition of the West?
WENGER: I have found going out and observing activities with my own eyes to be an incredibly useful thing. There are a number of aspects of malaria that maybe you know academically, but when you see them are in your face. My general experience of going around with the teams that are using these rapid diagnostics to test people and hand out drugs is that there are a lot of children between 5 and 10 years old who test positive and have no obvious symptoms, but are still capable of transmitting the disease. Another example would be the use of bed nets. One often sees bed nets that were handed out a couple of months previously that are still in their plastic packaging. And it's important to understand compliance with drugs. For instance, it makes a big difference on the impact of the program, and on the output of our model, whether the individual takes just one pill or takes the full regimen. So seeing what's on the ground has given me an important understanding of operational realities.
PT: Day to day, how do you spend your time?
WENGER: Split across a number of my responsibilities: Interacting with our various collaborators, running the simulations and analyzing their outputs, developing new features in the model, and coordinating the work of the various people here.
We are now a team of about 35 to 40 individuals. Some work on the core software, others work on the disease research on four different diseases.
PT: What is the funding model at the IDM?
WENGER: Our project is funded by Bill Gates. He's the primary investor in the Institute for Disease Modeling, and he meets regularly with our team to follow up on our progress.
PT: How does your physics background contribute to your work in disease modeling?
WENGER: There are three things: First, how you approach quantitative analysis. Approaching a problem with an unknown answer is something that coming out of an experimental physics background you almost take for granted—and don't know it's a skill until you realize how applicable it is in other fields. The second thing is software. My background in C++ reconstruction analysis software from the CMS [Compact Muon Spectrometer] experiment I worked on, modifying all the proton–proton algorithms to work in the context of lead–lead collisions is very applicable to the software effort here. The third thing is that particle physics is a very collaborative effort. The CMS experiment had, I think, 3000 members, which meant coordinating among the various different research groups, all of whom depended on the quality of the underlying reconstructed object, the tracking, the detector, and so on. And then there is communication between the theoretical and experimental sides of the field. That collaborative work really carries through to my current position.
Everyone brings his own experience to a problem, and that works really well when you have a couple of really different backgrounds. That's one of the things our group in the laboratory here takes really nice advantage of.