How Network Science Surfaced 81 Potential COVID-19 Therapies

Researchers led by Albert-László Barabási used network-based models to discover existing drugs that might take on COVID-19

3 min read
Proteins targeted by SARS-CoV2 are not distributed randomly in the human interactome, but form a large connected component (LCC) consisting of 208 proteins, as well as multiple small subgraphs.
Proteins targeted by SARS-CoV2 are not distributed randomly in the human interactome, but form a large connected component (LCC) consisting of 208 proteins, as well as multiple small subgraphs. Proteins not expressed in the lung are shown in orange, indicating that almost all proteins in SARS-CoV2 LCC are expressed in the lung, explaining the effectiveness of the virus in causing pulmonary infections.
Image: arXiv

Researchers have harnessed the computational tools of network science to generate a list of 81 drugs used for other diseases that show promise in treating COVID-19. Some are already familiar—including the malaria and lupus treatments chloroquine and hydroxychloroquine—while many others are new, with no known clinical trials under way.

Since the concept was first proposed in 2007, network medicine has applied the science of interconnected relationships among large groups to networks of genes, proteins, interactions, and other biomedical factors. Both Harvard and MIT Open Courseware today offer classes in network medicine, while cancer research in particular has experienced a proliferation of network medicine studies and experimental treatments.

Albert-László Barabási, distinguished university professor at Northeastern University in Boston, is generally considered the founder of both network medicine and modern network science. In a recent interview via email, Barabási said COVID-19 represents a tremendous opportunity for a still fledgling science.

“In many ways, the COVID offers a great test for us to marshal the set of highly predictive tools that we as a community [have developed] in the past two decades,” Barabási said.

Last month, Barabási and 10 coauthors from Northeastern, Harvard, and Brigham and Women’s Hospital in Boston published a preprint paper proposing a network medicine-based framework for repurposing drugs as COVID-19 therapies. The paper has not been submitted for peer review yet, says Deisy Morselli Gysi, a postdoctoral researcher at Northeastern’s Network Science Institute.

“The paper is not under review anywhere,” she said. “But we are planning, of course, to submit it once we have [laboratory] results.”

The 81 potential COVID-19 drugs their computational pipeline discovered, that is, are now being investigated in wet-lab studies.

The No. 1 COVID-19 drug their network-based models predicted was the AIDS-related protease inhibitor ritonavir. The U.S. Centers for Disease Control’s website lists 108 active or recruiting trials (as of May 6) involving ritonavir, with a number of the current trials being for COVID-19 or related conditions.

However, the second-ranked potential COVID-19 drug their models surfaced was the antibacterial and anti-tuberculosis drug isoniazid. ClinicalTrials, again as of 6 May, listed 65 active or recruiting studies for this drug—none of which, however, were for coronavirus. The third and fourth-ranked drugs (the antibiotic troleandomycin and cilostazol, a drug for strokes and heart conditions) also have no current coronavirus-related clinical trials, according to

Barabási said the group’s study took its lead from a massively collaborative paper from 27 March, which identified 26 of the 29 proteins that make up the SARS-CoV-2 coronavirus particle. The study then identified 332 human proteins that bind to those 26 coronavirus proteins.

Barabási, Gysi, and coresearchers then mapped those 332 proteins to the larger map of all human proteins and their interactions. This “interactome” (a molecular biology concept first proposed in 1999) tracks all possible interactions between proteins.

Of those 332 proteins that interact with the 26 known and studied coronavirus proteins, then, Barabási’s group found that 208 of them interact with one another. These 208 proteins form an interactive network, or what the group calls a “large connected component” (LCC). And a vast majority of these LCC proteins are expressed in the lung, which would explain why coronavirus manifests so frequently in the respiratory system: Coronavirus is made up of building blocks that each can chemically latch onto a network of interacting proteins, most of which are found in lung tissue.

However, the lung was not the only site in the body where Barabási and coauthors discovered coronavirus network-based activity. They also discovered several brain regions whose expressed proteins interact in large connected networks with coronavirus proteins. Meaning their model predicts coronavirus could manifest in brain tissue as well for some patients.

That’s important, Gysi said, because when their models made this prediction, no substantial reporting had yet emerged about neurological COVID-19 comorbidities. Today, though, it’s well-known that some patients experience a neurological-based loss of taste and smell, while others experience strokes at higher rates.

Brains and lungs aren’t the only possible hosts for the novel coronavirus. The group’s findings also indicate that coronavirus may manifest in some patients in reproductive organs, in the digestive system (colon, esophagus, pancreas), kidney, skin, and the spleen (which could relate to immune-system dysfunction seen in some patients).

Of course the first drug the FDA approved for emergency use specifically for COVID-19 is the protease inhibitor remdesivir. However Barabási and Gysi’s group did not surface that drug at all in their study.

This is for a good reason, Gysi explained. Remdesivir targets the SARS-CoV-2 virus specifically and not any interactions between the virus and the human body. So remdesivir would not have showed up on the map of their network science-based analysis, she said.

Barabási said his team is also investigating how network science can assist medical teams conducting contact tracing for COVID-19 patients.

“There is no question that the contact tracing algorithms will be network-science based,” Barabási said.

The Conversation (0)

Why Functional Programming Should Be the Future of Software Development

It’s hard to learn, but your code will produce fewer nasty surprises

11 min read
A plate of spaghetti made from code
Shira Inbar

You’d expectthe longest and most costly phase in the lifecycle of a software product to be the initial development of the system, when all those great features are first imagined and then created. In fact, the hardest part comes later, during the maintenance phase. That’s when programmers pay the price for the shortcuts they took during development.

So why did they take shortcuts? Maybe they didn’t realize that they were cutting any corners. Only when their code was deployed and exercised by a lot of users did its hidden flaws come to light. And maybe the developers were rushed. Time-to-market pressures would almost guarantee that their software will contain more bugs than it would otherwise.

Keep Reading ↓Show less