There are still more questions than answers. For the first months of the COVID-19 pandemic that swept across the globe in the spring of 2020, doctors and researchers were mostly stumbling in the dark as they fought a new and sometimes deadly infectious disease. One of the most conspicuous symptoms among patients infected with SARS-CoV2, the novel coronavirus that causes COVID-19, was respiratory distress: difficulty breathing, pain or pressure in the chest, and CT scans that revealed what were described as “ground-glass opacities,” or abnormally hazy areas, in patients’ lungs. For this reason, an early – and incomplete – concept of COVID-19 emerged: It was a respiratory disease.

Over time, a more complex and bewildering picture has developed. COVID-19 isn’t limited to the respiratory system. Patients with the disease experience a variety of symptoms that can include muscle and joint pain, extreme fatigue, cardiac arrythmias, blood clotting, skin rashes, nausea, vomiting, diarrhea, headaches, and decreased cognitive functioning or “brain fog.” In the spring and summer of 2020, several reports of younger COVID-19 patients with a mysterious and dangerous array of symptoms, known as multisystem inflammatory syndrome, emerged.

Equally puzzling are the variations in how the disease affects infected people: A large percentage – according to the U.S. Centers for Disease Control and Prevention (CDC), up to 70 percent, with a most recent “best estimate” of 40 percent – experience no symptoms at all. Many patients experience it as a mild upper respiratory infection, with a fever and slight cough; many suffer more severe symptoms, often in clusters; and a significant number of people are killed by COVID-19. The high percentage of asymptomatic patients makes it difficult to estimate an accurate infection fatality rate for the disease, but the number of deaths so far – by October 2020, more than a million people worldwide, including more than 220,000 Americans – is far greater than the annual deaths associated with seasonal influenza.

Why such a wide array of symptoms? Why such variation in the complexity and severity of symptoms? Why such a vast difference in patient outcomes? In April 2020, it began to look as if disease experts were approaching some answers: Intriguing results began to emerge from studies conducted by a team of clinicians and systems biologists who had fed patient data from publicly available datasets into supercomputers at the Department of Energy’s (DOE) Oak Ridge National Laboratory. This group’s findings have offered insights into disease pathways in the body, possible explanations for the multitude of COVID-19 symptoms, and – most important – ideas for how existing drug therapies might be adapted to counteract these symptoms.

The Oak Ridge team was led by computational systems biologist Daniel Jacobson, PhD, with clinical insights supplied by doctors including Amy Justice, MD, PhD, a staff physician and clinical epidemiologist at the Department of Veterans Affairs (VA) Connecticut Healthcare System and professor of medicine and public health at the Yale School of Medicine. The team began by analyzing gene expression patterns in the lung fluid cells of COVID-19 patients and, using the lab’s Summit supercomputer, comparing them with the genes of uninfected patients. These computational analyses suggested that genes involved in regulating the release of bradykinin – a peptide hormone that promotes inflammation, dilating blood vessels to lower blood pressure – appear to be excessively “switched on” in the lung cells of those with the virus.

The resulting “bradykinin storm,” if it occurred in other organs, could explain the wide variety of symptoms experienced by COVID-19 patients. The team promptly turned its attention to how the bradykinin system and the renin-angiotensin system (RAS) – the peptide system involved in regulating blood pressure, degrading bradykinin, and contributing to other imbalances – react to SARS-CoV2 infection. The follow-up study was a population-scale deep dive, comparing gene expression in 57 different tissues in an uninfected group of 1,300 people – a study of 17,000 RNA-sequence datasets overall.

“There’s just a really good matchup between what we would predict from the bradykinin in different tissues and the range of symptoms we see in COVID-19, which everybody has been scratching their heads about,” said Jacobson. “How does this virus give us this whole collection of symptoms that are not typical of most viruses, and why do some people have one symptom and not others?” One explanation, he said, may be that viral colonization of tissues are localized; infection may take different routes in different people. “At the population scale, we were able to show there’s great variance in the expression of these key genes across the population,” Jacobson said. “So that’s probably a combination of underlying genetic variance, as well as environmental effects.”

A NOT-SO-NEW PARTNERSHIP

Armed with a molecular model for how the RAS-bradykinin pathway responds to the novel coronavirus, and with the computing power to apply it to different populations and scenarios, researchers are poised to investigate a multitude of questions. If the model proves accurate, it can provide a template for how, and when, the disease can be treated, perhaps with existing U.S. Food and Drug Administration (FDA)-approved drugs. In their report on the study, Jacobson’s and Justice’s team identified drugs or supplements that might be used to treat COVID- 19: medications such as stanozolol or ecallantide that reduce bradykinin production; bradykinin receptor inhibitors, such as icatibant; and vitamin D, which is involved in the RAS system and could attenuate or prevent a bradykinin storm.

Several studies are already underway within the COVID-19 Insights Partnership, a framework launched in April 2020. The new initiative joins the DOE’s computing power with the federal agencies overseeing the largest repositories of health data in the nation: The VA and the Department of Health and Human Services (HHS).

Collaboration between the DOE and VA isn’t new. Justice’s participation in the Oak Ridge project was established, in part, because discussions about forming the COVID-19 Insights Partnership had begun months earlier, when VA’s chief research and development officer, Rachel Ramoni, PhD, floated the idea of combining federal health data resources with DOE’s high-performance computing and artificial intelligence capabilities to explore questions about the COVID-19 pandemic. VA’s Office of Research and Development, the FDA, HHS, and DOE, recognizing the enormous potential of this data to study and perhaps solve some of the mysteries of COVID-19, began discussing a partnership in March 2020, when U.S. cases numbered in the hundreds.

The speed with which the new partnership was launched is due in part to the fact that VA electronic health record data was already mirrored on servers at Oak Ridge: Four years ago, the two agencies launched an initiative, MVP-CHAMPION, to apply DOE’s supercomputing, artificial intelligence, and data analytics to VA’s unparalleled health care data and genomic data compiled by the VA’s Million Veteran Program (MVP). Since 2017, computing and clinical experts have been working to gain insights into three topics important to veteran health: cardiovascular disease, cancer, and suicide. The outcomes of these ongoing projects will provide better predictors of disease and improved clinical decision support for VA care providers and patients.

As Justice pointed out, moving VA data onto DOE servers required a lot of work to get all the permissions and privacy protections in place. “All of that happened before COVID-19 hit,” she said. “If it hadn’t, none of this would have happened.” Justice compared the unexpected benefit of the MVP-CHAMPION initiative to NASA’s Apollo program, which ushered in a sequence of unforeseen technological innovations. “When COVID-19 hit,” she said, “all of a sudden there was an immediate use for that data that no one could have anticipated when we first started the collaboration.”

According to Molly Klote, MD, who directs the VA’s Office of Research Protections, Policy, and Education, MVP- CHAMPION was formed for the benefit of both agencies: “The Department of Energy needed to practice modeling with the use of their supercomputers, and they needed an enormous data set,” she said. Genomic data from the MVP – obtained so far from about 500,000 of the 830,000 veterans participating in the program – comprises the largest such database in the United States.

It’s been Klote’s job to navigate a new set of regulatory and privacy concerns, to unlock this data trove for analysis of COVID-19. “The enclave at the Department of Energy where all the VA data sits is a VA enclave that meets all of VA’s standards for privacy and security,” she said. Disclosure of a veteran patient’s data requires a series of permissions, usually including the signature of the veteran, but because the COVID-19 pandemic is a public health emergency, the Partnership has received permission under HHS Public Health Authority to pool health record data from the VA and the Centers for Medicare and Medicaid Services (CMS).

Because the RAS-bradykinin computational study was formulated while Klote and her office were securing this public health authority, veteran data was not used to develop the mechanistic model developed at Oak Ridge – but as Jacobson pointed out, the COVID-19 Insights Partnership is about more than sharing data; it’s also about sharing expertise. “The partnership with Amy [Justice], and getting her perspectives on the mechanism, was extremely helpful,” he said. “This was a good outcome of the Insights collaboration, and it’s leading to a lot of what we hope will be very productive interactions with the VA.”

 

NEXT STEPS

Studies are underway both within and outside of the COVID-19 Insights Partnership. Three studies of existing drugs, each targeting a different part of the RAS-bradykinin pathway – icatibant; dexamethasone; and calcifediol, a vitamin D analog – have yielded early results that support the team’s disease model, and the partnership is working to facilitate clinical trials of other drugs identified by the Oak Ridge team. According to Justice and Jacobson, it’s likely that therapies for COVID-19 will combine two or more of these drugs, in courses targeting the needs of individual patients, to shut down the mechanisms of COVID-19 pathogenesis.

The COVID-19 Insights Partnership’s earliest studies of VA and CMS data have been straightforward analyses of questions posed by the FDA, many of which don’t require complex algorithms and artificial intelligence: Some have focused on operational questions (i.e., the usage rate of existing drugs), while some examine treatment outcomes (the safety and efficacy of drugs such as anticoagulants and hydroxychloroquine).

According to Justice, priorities among the list of questions being explored by the Insights Partnership are adjusted as new findings are revealed. “We have agreed that this group, the Insights group, can arbitrate what questions will be addressed and how they will be prioritized,” she said. “For example, hydroxychloroquine initially was on the list of things they wanted to look at. Then, over time, the antithrombotics became very important, because of all the thrombosis that we’re seeing. So the group asked to add antithrombotics to the list. We’ve completed an analysis of hydroxychloroquine, which we fed into a larger collaborative analysis of several other groups involved with the FDA.”

These analyses, said Justice, have focused exclusively on data from the VA, which, as one of the nation’s largest health care systems, also has one of the largest samples of COVID-19 patients – as of October 2020, the VA health care system had treated more than 65,000 people for the disease. Much work remains to be done; these early studies have been mostly limited to data that could be extracted and studied by Justice’s group from VA records. Over time, as DOE frees up more resources – people and time – to devote to the Partnership, it will enable more ambitious computational studies.

As the partnership grows, said Klote, and new data sets are added to the sample for analysis, it will be important to focus on the diversity of the data: Despite their immensity, the VA and CMS data sets represent a pretty narrow sample of Americans. “About 60 percent of our population is Vietnam veterans, almost exclusively men,” she said. “We just don’t have a very diverse population. So we can say certain things about the data, but we’re limited, in a lot of ways, about what we can say. That’s why we were trying to pull all these different data sources onto the Department of Energy servers.” Klote is enthusiastic about the prospect of accessing data from the military health care system, which treats not only active-duty service members but also their spouses and kids, as well as military retirees. “We don’t have any kind of population like that in VA,” she said.

Amid the worst pandemic the world has seen in a century, the architects of the COVID-19 Insights Partnership are focused on throwing everything they have at increasing our understanding of the disease and improving outcomes for patients. “One of the things we’re learning by trying to do this collaboration is how can we most effectively do it in a way that is respectful, protecting individual privacy, but also capitalizing on the fact that we have this information that can help both veterans and non-veterans,” said Justice. “It’s been a steep learning curve, but we’re hoping we’re going to be better positioned to be able to do this even more effectively in the future, particularly if there is another wave of the COVID-19 epidemic.”

Klote, who spent 13 years overseeing Army medical research programs before joining the VA, believes the COVID-19 Insights Partnership is a groundbreaking collaboration, with the potential to impact the future of American medicine. “If we can put all these data in the same place, it will be a model for future public health emergencies, such as suicide or opioid addiction,” she said. “If we can get agreements together and work through this process – because it’s never been done before – we will have a model to solve some major medical issues going forward.”

Click here, to get your free print edition of 
Veterans Affairs & Military Medicine OUTLOOK
delivered to your home or office.

---