Pandemic preparedness

During the pandemic turmoil, Margaret Cheung reconsidered her career. At the University of Houston, she was approaching the physics of proteins theoretically, such as how they fold and interact, but the pandemic changed her life.

“Most of my students and researchers were working from home, and I had young children at that time who had to learn from home,” she says. “I realized that it was a big disruption, and it prompted me to rethink my career priorities.”

She picked biopreparedness and, in 2021, joined the Department of Energy’s Pacific Northwest National Laboratory (PNNL) as a computational scientist. “All of my research has been sharply realigned to understanding how things work inside the cells,” Cheung recalls. “My research was on that path before, but it was more of a science pursuit. Now, it’s a mission.”

Cheung received one of the 10 awards by DOE Biological and Environmental Research program’s Biopreparedness Research Virtual Environment (BRaVE).  Its aim: to model the molecular mechanisms of disease transmission. To do that, she is developing a digital twin of the process.

 

“A digital twin is all about improving the efficiency of a workflow,” she explains. As an everyday example, she mentions Google Maps, a digital twin of the walkways, streets, roads and highways people travel. With a person’s current location and desired destination, Maps suggests alternative routes to get around closed roads or traffic jams to save travel time.

Cheung has applied a similar approach to biopreparedness. If a pathogen infects a cell or a host organism, a digital twin of the disease transmission can explore pathways that might stop the infection’s spread. “That’s how we help the scientists develop better strategies to improve human health and human quality of life,” she says.

Developing a digital twin starts with a cell. “A cell is a factory,” Cheung says. “It’s an engine.” By collecting information on how a virus alters a cell’s molecular processes, scientists might develop, for example, vaccines that disrupt the viral mechanisms. But that takes an enormous amount of data from decades of research in biology, virology, epidemiology and so on.

“The biggest challenge is understanding all of the data because biological data is so vast and messy,” Cheung says. “But we want to integrate all of that.” A digital twin might tease out a mechanism of viral infection and integrate that information with wet-lab experiments. With that gigantic database of information and a validated model, though, a digital twin could tweak the complex pathways and see if that reduces threats to human health.

“That is a very complex workflow,” Cheung says. “It’s challenging to figure out exactly where a process has happened, but this challenge really points out the urgent need to develop a digital twin.”

To develop a dataset for a biopreparedness digital twin, Cheung and her colleagues focus on host interactions with rapidly evolving pathogens. Then they’ll collect as much data as possible about the host, pathogen and what happens when they run into each other.

‘This data is diverse. The collection and analysis generate even more data.’

For example, collecting this information includes sequencing the pathogen’s genome. Combining a pathogen’s genomics with its proteomic information — the proteins that an organism produces — can track how the organism changes over time or in different places. Pathogen images via electron microscopy can reveal changes in shape and structure. Environmental and chemical data about where a pathogen and host interact will also be collected. Plus, data from a pathogen and host can be combined and compared.

“This data is diverse,” Cheung says. “The collection and analysis generate even more data.”

A biopreparedness digital twin isn’t just for human diseases. Cheung and her colleagues also will be studying environmental dangers, such as those faced by Prochlorococcus marinus, a cyanobacteria species that plays a crucial role in capturing climate-warming carbon dioxide. All the while, scientists must keep in mind that viral infection of these cyanobacteria can reduce their carbon-capturing capabilities or even kill them. Alternatively, the virus and cyanobacteria can co-evolve and survive environmental stresses. P. marinus is just one small part of the biopreparedness puzzle, which gives a glimpse of the expansive challenge of modeling biopreparedness in general.

Cheung and her PNNL colleagues are using a current ASCR Leadership Computing Challenge (ALCC)  award to further explore these processes. This award provides time on several high-performance computers: Aurora and Polaris at the Argonne Leadership Computing Facility, Frontier at the Oak Ridge Leadership Computing Facility and Perlmutter at the National Energy Research Scientific Computing Center. The power of these platforms ranges from peta- to exascale — the ability to perform a thousand-trillion to a million-trillion operations per second, respectively. The variety of platforms is intended to develop a digital twin that can run on a range of architectures.

The biopreparedness twin consists of a gigantic number of parts, from biological and environmental data to mathematical models and computational algorithms, but the key lies in creating an easy-to-use and powerful tool.

“This twin is for predicting what we could do based on learning from past data — not just learning but making sense of it,” Cheung emphasizes. “That requires a real model that puts these pieces together, so that we can understand what it means.”

Cheung and her colleagues, though, will not be working alone. The other nine BRaVE projects taking place at other national laboratories can eventually be connected.

“BRaVE is really Manhattan Project 2.0, involving multiple national labs, with the vision of creating a virtual laboratory across all of us,” Cheung says. “Hopefully that happens very quickly because there’s some urgency to integrate all these things, before the next pandemic or biological disaster.”

Although Cheung’s team is just getting organized to use the ALCC award, she’s focusing on a couple of goals. The first is simply using the allocation to compute the data. Second, she wants to train computational scientists to run the digital twin.

“A lot of us are biological or environmental research scientists, and very few of us have extensive experience in running simulations,” Cheung says. As a physicist, though, she does have that experience and hopes to pass it along to others. “Once we train them, I would expect new generations of computational scientists at PNNL would be more comfortable using it.”