From Terrorist Cells to Sexual Networks: Modeling Complex Systems

Graphic Network of People


Fresh out of the University of Pennsylvania, Ravi Goyal was recruited by the National Security Agency (NSA) for its rapidly expanding Counter-Terrorism Unit. The year was 2002. Ravi went to work as an Applied Research Mathematician, using his exceptional math and computer skills to analyze how information flowed within and between terrorist networks.

But it wasn’t all sitting at a desk. During the Iraq War, he was embedded in an intelligence unit in Baghdad, assigned to find new ways to utilize information. “It was important to see not just data, but how it was coming in,” said Ravi. “That really impacts how you view it and how you want to use it and the uncertainties and errors that come with various kinds of collections.”

When Ravi’s wife, a nurse, suggested he could use his skills to make a positive impact in the field of public health, he took her advice. He left the NSA and enrolled in a PhD program at the Harvard School of Public Health (HSPH). These days, rather than studying how information moves between terrorists, he looks at how HIV spreads within communities.

How to Build a Model

In the Botswana Combination Prevention Project (BCPP), researchers led by Max Essex are conducting a large clinical trial to determine if combining several HIV prevention measures can significantly reduce the number of new infections within a community.

As part of the project, Ravi and his colleagues are building a computer model that creates a simulation of how the AIDS virus spreads within a village. This new model, one of the innovative aspects of the BCPP, will allow researchers to measure how effectively their interventions are working.

Gathering information is the first step in building a model. In the BCPP, information will be collected from tens of thousands of study participants over the course of several years, resulting in billions of bits of data. The data comes from a number of sources, including extensive testing for HIV and lab tests from people who are infected.

Dynamic evolution of the sexual contact network: Individuals are represented by circles and a line connecting two individuals represents a sexual relationship between the individuals.
Dynamic evolution of the sexual contact network: Individuals are represented by circles and a line connecting two individuals represents a sexual relationship between the individuals.


Demographics, such as the fact that about 25% of the adult population of Botswana has HIV, are fed into the model, as is knowledge gained from previous HAI clinical trials like the Mochudi Prevention Project.

Most new HIV infections in Botswana are from sexual contact. Though researchers can’t actually observe the sexual network, they use several methods to get a glimpse of it. Questionnaires from study participants provide detailed information, such as number of sexual partners, length of relationships, and knowledge of partner’s HIV status. Participation in the study is voluntary and all information is confidential.

As enormous amounts of data are collected, the model gives researchers a way to organize and interpret it. Members of the BCPP biostatistics team, led by Dr. Victor DeGruttola, use algorithms and mathematical techniques to simulate the complex biological evolution of a disease within an individual, while simultaneously modeling how the HIV virus moves from one person to another along an evolving sexual network.

The BCPP model must deal with a constantly changing situation on the ground. In the clinical trial, two processes are operating over time. One is the spread of new HIV infections throughout the community; the other is the roll-out of HIV prevention interventions within the community.

As the trial unfolds, more and more people will be tested for HIV. Testing will get more people on antiretroviral treatment (ART). Once an individual is on treatment, his or her viral load drops, as does the chance of transmitting HIV to a partner.

Along with increased testing and treatment, there will be a roll-out of voluntary male circumcision, which has been shown to reduce transmission of HIV by over 50%. Condom distribution and other prevention measures will also be increased.

As information is added, the model evolves. It will be able to provide preliminary assessments about the effectiveness of the clinical trial. “We want to know early on whether or not the trial looks as though it’s succeeding,” said DeGruttola. “The model will allow the team to make mid-course corrections if necessary.”

Researchers are also using a new application of genetic sequencing—viral genetic sequencing of new infections—to learn how closely related new infections are to other infections within a community. This is somewhat like a paternity test for HIV, showing whether new infections are related to known infections. Viral genetic sequencing will not only provide information for the model, it will also serve as a check to measure how accurately the model reflects what’s actually happening on the ground.

Conducting the BCPP on the ground in Botswana is an enormous scientific and logistical challenge. The clinical trial involves tens of thousands of study participants and hundreds of doctors, nurses, counselors, and staff to follow them over the course of several years. Using the model, researchers can explore a number of options that would be too expensive to test in the real world. “It wouldn’t be practical or ethical to run a whole bunch of experiments over and over again on a group of people,” said Ravi, “but a model allows us to test different kinds of interventions and how they would work under various conditions.”

The results of the BCPP could provide the first real example of how to systematically end the AIDS epidemic in Africa. The state-of-the art computer model being developed will be adaptable for use in other countries with different types of AIDS epidemics.
Ravi is just finishing his PhD and will stay on as a postdoc. It turns out that working on the AIDS epidemic is not entirely different from working on terrorist networks. “You want to see how they evolve,” said Ravi, “and you want to see how you can break them apart.”