COVID-19’s spread is roaring back to life across the U.S. as widespread reopenings take their toll, giving a renewed urgency to the scientific research frantically building a better understanding of the virus and searching for therapeutics and vaccines. Some of these researchers – a team from the University of California, San Diego and Maynooth University in Dublin – used supercomputing power to zero in on one element of the virus: its deceptive sugar coating.
Fittingly, COVID-19’s sugar coating effectively sweet-talks the human body. The virus’ spike proteins – the battering rams it uses to enter human cells and duplicate itself – are coated with glycans, which the human immune system sees as harmless. These glycans are so substantive that they account for around 40 percent of the spike proteins’ weight.
“You really see how effective its glycan shield is,” said Rommie Amaro, professor of chemistry and biochemistry at the University of California, San Diego, in an interview with Jorge Salazar at the Texas Advanced Computing Center (TACC). “That’s because you see the glycans covering the surface of the viral spike protein, which is the most exposed bit and the part that’s responsible for the initial infection in the human cell.”
Amaro and her team simulated the spike protein and its glycans, rendering the dynamics of 1.7 million atoms – a task impossible for anything but a supercomputer. For that, they turned to the Frontera system at TACC.
Frontera’s 8,008 compute nodes are equipped with Intel Xeon CPUs, 192 GB of memory and a Mellanox InfiniBand HDR100 interconnect, and its two subsystems are equipped with four Nvidia GPUs per nodes (one subsystem uses Quadro RTX 5000s, while the other uses V100s). Overall, Frontera delivers 23.5 Linpack petaflops, placing it fifth on the most recent Top500 list of the world’s most powerful publicly ranked computers.
The research team used up to 4,000 nodes – nearly half – of Frontera, racking up 2.3 million node hours in compute time. “The reason why the computer resources at TACC are so important is that we can’t understand what these glycans look like if we don’t use simulation,” Amaro said.
These simulations illuminated some surprising behavior in the spike proteins and their glycans. When the spike protein enters the human cell, it contorts itself significantly, and that contortion exposes parts of the spike protein.
“When that receptor binding domain lifts up into the open conformation, it actually lifts the important bits of the protein up over the glycan shield,” Amaro said. “Our analysis gives a potential reason why it does have to undergo these conformational changes, because if it just stays in the down position those glycans are basically going to block the binding from actually happening.”
This, Amaro explained, is the virus’ most dangerous state: when the spike exits the glycan shield, she said, it is “locked and loaded.”
“Now we’re trying to share our data with as many people as we can,” Amaro said. “Because people want a dynamical understanding of what’s happening – not only with other academic groups but also with different pharmaceutical and biotech companies that are conducting neutralizing antibody development.”
To read the article discussing this research from TACC’s Jorge Salazar, click here.
Header image: The glycan shield (dark blue) protects the spike protein as it prepares to enter a human cell. Image courtesy of the researchers.