One of the very few serious side effects to emerge from any of the (broadly miraculous) Covid-19 vaccines has been the emergence of blood clotting, common only among adenovirus-based vaccines like those offered by AstraZeneca (United Kingdom), Johnson & Johnson (United States) and Sputnik (Russia). With vaccine skepticism plaguing many areas of the world, understanding and preventing the mechanisms that cause such adverse side effects is crucial to advancing vaccination efforts. Now, researchers from Arizona State University (ASU), the Mayo Clinic and AstraZeneca (among others) have collaborated on supercomputer-powered research that may have unearthed a causal factor behind the blood clots associated with the adenovirus-based Covid vaccines.
“[This side effect is] a very rare disorder, now, of course,” Abhishek Singharoy, an assistant professor in the molecular sciences at Arizona State University (ASU), explained in an interview with the Extreme Science and Engineering Discovery Environment (XSEDE). “But … when you tally the numbers of the vaccines … given all around the world, including in the U.S., that factors to a big number of potential patients. That’s why it remains important.”
Prior to this research, other scientists had noted a bloodstream-based protein associated with clot formation—”platelet factor four,” or PF4—adhering to particles of the vaccines in question. These researchers built on that discovery with a method called Brownian dynamic simulation, which uses first principles to predict interactions between molecular structures.
To run these intensive simulations, the researchers sought an XSEDE supercomputer allocation, which they obtained for the Bridges-2 supercomputer at the Pittsburgh Supercomputing Center (PSC). Specifically, the research employed Bridges-2’s “regular memory” nodes; the system has 488 of these, each with dual AMD Epyc 7742 CPUs and at least 256GB of memory (16 of the nodes have 512GB each).
“We wanted statistically significant results, so we actually ended up running 400 parallel simulations, hundreds of parallel computations of the molecular docking of PF4 on the surface of the adenovirus vector,” Singharoy said. “Now we’ve seen in molecular detail the structure of the vaccine and realized that the outer surface … is negatively charged, which was not something that you could guess [without the simulations].”
The simulations, it turned out, were enormously helpful: they showed that PF4, which usually has a positive charge, was sticking to the negatively charged vaccine particles. When subsequent simulations incorporated the anti-clotting drug heparin, it prevented this sticking behavior. Wet-lab analysis bore out the simulation results.
Still, this remarkable research does not offer all the answers—only a promising glimpse at the beginning of the clotting process. Questions remain: why is clotting only observed in a small percent of those vaccinated? How can the adenovirus-based vaccines be reengineered to prevent the clotting?
Next, the researchers plan to employ more supercomputing on Bridges-2—including AI-accelerated work—to help answer these questions.
To learn more, read the reporting from XSEDE here.