Modeling to control infectious diseases

Pathogens generate a complex set of signals as they spread through human populations. A single infection can result in a set of clinical symptoms, a case report in a hospital surveillance system, a genetic sequence, a treatment outcome, an antibody repertoire, and modified social behavior.

The Boni Lab investigates these different data points and data streams with methods at the interface of field, clinical, and computational epidemiology. From 2008 to 2016, we were based at the Oxford University Clinical Research Unit in Ho Chi Minh City. Currently, we are based at the Center for Infectious Disease Dynamics, Department of Biology, Pennsylvania State University.

See our pages on SARS-CoV-2 related work.

Current Collaborations

Malaria Modeling Consortium: We are part of the Bill and Melinda Gates Foundation’s Malaria Modeling Consortium (MMC) which currently includes several of the world’s leading mathematical modeling groups in malaria. Our current role is to lead Work Package 2 on drug-resistance evolution, whose main goal is to evaluate stewardship plans for ACTs and other novel antimalarials over the next decade.
OUCRU and NIH-CEIRS: We recently received funding from the Johns Hopkins Center for Excellence in Influenza Research and Surveillance (CEIRS) to investigate how antibody profiles change after an influenza infection, and whether broadly-neutralizing antibody to the HA2-stalk region is sufficiently long-lasting for vaccination purposes. This project will be carried out in collaboration with the Oxford University Clinical Research Unit in Vietnam.
Arboviral Interactions in Panama: In 2019, we began working with the Instituto Conmerativo Gorgas de Estudios de la Salud to measure interactions rates among a range of arboviruses — dengue virus, Zika virus, chikungunya virus — that are endemic to Central and South America. Our current role is to reconstruct about 10 years of arboviral epidemiology in Panama to determine if different viral epidemics interfere with each other.
DeTACT: In 2019, the DeTACT trial — "Developing Triple Artemisinin Combination Therapies" — began recruiting patients in Asia and Africa. The trial is led by PI Arjen Dondorp from the Mahidol-Oxford Research Unit in Bangkok. We are leading the mathematical modeling work package whose aim is to assess the risks of drug resistance emerging to triple ACTs and optimal population-level deployment strategies once TACTs are approved.

Research Areas

Influenza in the Tropics: The majority of our research questions fall under the umbrella of tropical influenza epidemiology. We look at the circulation of human or “seasonal” influenza viruses in the tropics as well as the ecology and evolution of avian influenza viruses and factors that are linked with human exposure. We run both field and clinical studies in southern Vietnam. Our laboratory methods focus on identifying antibody repertoires and viral sequences, and our analytical methods are rooted in mathematical modeling and fitting models to data with likelihood methods. Our key goals are to characterize influenza’s persistence patterns, its seasonal patterns (or lack thereof), the relationship between influenza circulation and other respiratory viruses, and the symptomatic/asymptomatic nature of influenza epidemics in Vietnam.
Multiple First-line Therapies for Malaria: The area of the most direct public health relevance that we work in is analysis and optimization of population-level treatment strategies for malaria. A balance must be struck when designing a population-level treatment strategy, as high levels of treatment drive drug resistance evolution but low levels are associated with high morbidity and mortality. One method of lowering the risk of resistance evolution to an individual drug, while maintaining high treatment rates in the population as a whole, is to deploy multiple drugs simultaneously in the population. Our epidemiological microsimulations have shown that recommending simultaneous use of multiple first-line therapies (MFT) for malaria is a better public health strategy that the status quo approach.
Big-data Seroepidemiology: Since the 2009 influenza pandemic, repeated cross-sectional seroepidemiological study designs have become a common way to explore the dynamics of infection, susceptibility, and post-infection antibody responses. In southern Vietnam, we have initiated a study in which periodic collections of population-representative serum are collected every few months, and we are currently testing for the presence of influenza antibody in these samples. The resulting data set is structured as a serological time series, or an antibody time series, and it allows for the inference of past disease dynamics if the underlying epidemiological models of infleunza transmission are believed to be correct. The inferential process reconstructs complete disease dynamics, i.e. the dynamics of symptomatic, subclinical, and asymptomatic influenza infections.
Participatory Epidemiology: One of our key study frameworks is a community network of general practitioners in Ho Chi Minh City who send out daily reports of case numbers of influenza-like illness (ILI) by standard SMS text messages. These ILI reports are aggregated daily and reported in real-time at www.ili.vn to show the current and recent trends of ILI activity in Ho Chi Minh City. One of the key questions we will be answering in this study is whether ILI trends and influenza trends are correlated in the tropics. If they are not, influenza surveillance systems in the tropics will need to place an emphasis on virological/molecular confirmation over syndromic surveillance. In addition, other repiratory viruses will need to be studied in more detail in Vietnam to determine the major causes of non-influenza ILI peaks.
Recombination Detection: Part of our work is focused in bioinformatics, and the key tool that we have been maintaining for ten years is the recombination detection algorithm 3SEQ along with an online statistical calculator that can be used more generally to test the null hypothesis of “no mosaicism” in any type of string or sequence. The statistical test used in this tool is a non-parametric test for detecting clustering in one dimension (also called anomalous interval detection). It simply detects if one set of binary observations is clustered in the middle of second set of binary observations, and it can be viewed as a two-breakpoint version of the Mann-Whitney U-test or the Wilcoxon Rank-Sum test.
Epidemiological Theory: All of our work is founded in epidemiological theory. In addition to the field, clinical, and sequence based work that we do, we try to keep up with and contribute to the literature on study design, pathogen ecology and evolution, evolutionary epidemiology, and economic epidemiology. Our malaria work is rooted in the theory that evolutionary adaptation is slow in variable environments. Presenting the malaria parasite with simultaneous multiple lethal challenges should make it difficult for the parasite to evolve resistance to all of them. Likewise, part of our influenza work seeks to test the theory that long chains of influenza transmission in the tropics afford the virus more opportunity to accumulate the necessary genetic mutations to escape population immunity through antigenic drift. Testing this hypothesis requires a time series of population immunity, virus sequences, and case numbers – something we have been working towards over the past five years.