BU Interdisciplinary Team Wins DOE Grant for Biopreparedness Research

By Chloe Patel, Hariri Institute for Computing

Boston University researchers affiliated with the Hariri Institute, College of Engineering, and School of Public Health, have received a subaward from a $12 million U.S. Department of Energy (DOE) grant awarded to a coalition of DOE National Laboratories led by the Lawrence Berkeley National Laboratory (LBNL). This funding, provided by the Office of Science, will support fundamental research to accelerate breakthroughs in support of the Biopreparedness Research Virtual Environment (BRaVE) initiative.

The LBNL-led project aims to develop generalized agent-based models (ABMs) for diseases and epidemiology, and model a variety of time-evolving diseases. The goal is to create new workflows that incorporate both reinforcement learning and surrogate models to optimally evaluate a variety of intervention scenarios and generate forecasts, with uncertainties, to guide policy decisions. In addition to Boston University, the LBNL-led project includes a consortium of DOE national labs, including Argonne, Brookhaven, Los Alamos, Livermore, and Sandia National Laboratories.

“The COVID-19 pandemic has revealed significant gaps in our ability to forecast, model, and more so in deciding on specific control measures that can be deployed,”  says BU PI Yannis Paschalidis, distinguished professor of engineering (ECE, SE, BME), founding professor of Computing & Data Sciences, and director of the Hariri Institute. “This project will leverage DOE’s ABM capabilities, coupled with exascale computing, to develop improved epidemic spread models, calibrate them for various key pathogens with pandemic potential, develop uncertainty quantification capabilities, as well as methods for early detection and mitigation of an emerging epidemic.”

An ABM is a computational model used to simulate interactions between “agents” which could be people, places, things, or time. These agents allow a more dynamic approach to modeling.

 “The idea is literally to have a model of a city with a little entry of every person in the city,” says Co-PI Alex Olshevsky, Hariri Institute Faculty affiliate  and associate professor of electrical and computer engineering (ECE). “So you model people as going to work, you model people as going to the grocery store, and you run a basic simulation to see how the disease progresses, depending on what you do.”

The simulations, through trial-and-error, will attempt to determine the best way to “reduce disease spread, while causing as little economic pain as possible, according to Olshevsky. 

In addition to Paschalidis and Olshevsky, the BU team includes Helen Jenkins, an associate professor of biostatistics in the BU School of Public Health and a former Hariri Institute Junior Faculty Fellow. 

The three-year grant allows the BU researchers  to contribute in two ways, by first detecting local disease clusters and then to develop societal mitigation strategies. To detect disease clusters, they will develop new machine learning algorithms that leverage electronic health records and other streams of data to detect non-typical disease patterns, spikes in the number of people seeking treatment, or unexpected outcomes for the hospitalized — for example, excess ICU admissions or deaths. They will also develop reinforcement learning methods to formulate more adaptive and dynamic intervention policies, from localized travel restrictions, targeted mask-wearing, selective quarantining, and testing, all of which are essential for policy decision making in fast-moving, complex epidemics.

The researchers will also develop methods to allow for more adaptive and dynamic intervention policies, which is essential for policymakers’ decision making in fast-moving, complex pandemic.

For their work, BU researchers will use diverse sources of data, either de-identified Electronic Health Records (EHRs) in healthcare centers of developed parts of the world, or, in the case of less well-resourced settings, other forms of data ranging from local case reports, social media, web search queries, cell phone data, and death reports.

In 2022, Paschalidis led a BU team which received a grant from the National Science Foundation entitled, Predicting and Preventing Epidemic to Pandemic Transitions, which looked to find a natural progression from prediction and detection to prevention.

 Similarly, with this research, Paschalidis, Olshevsky, and Jenkins will also look to recommend specific control mechanisms for prevention through both model predictive control and reinforcement learning. 

“BU has a large team in place working on predicting and preventing the next pandemic. So we expect that this experience will help the new project,” says Paschalidis. “We are excited to join a large consortium of DOE national labs to help take DOE’s superb capabilities in ABMs to the next level, developing models that would be extremely useful in future pandemic prediction, characterization, and mitigation.”