During the pandemic, protein binding simulations have had a heyday as researchers scrambled to use supercomputing to find effective ways to inhibit SARS-CoV-2’s spike protein. But developing a better understanding of protein binding has a wide range of applications outside of treating COVID-19, ranging from drug screenings for a wide range of diseases and conditions to the development of novel biological materials. Now, researchers have leveraged the Summit supercomputer at Oak Ridge National Laboratory (ORNL) to study protein-protein binding, shedding new light on how those processes function.
Specifically, the researchers set out to isolate the role of shape in protein-protein binding processes, looking to see if they could predict binding based on the shapes of the ligand and the receptor alone. To do this, they simulated the assembly of 46 protein pairs with known affinities.
The simulations were conducted on ORNL’s Summit supercomputer, which, at 148.6 Linpack petaflops, remains the second most powerful publicly ranked supercomputer in the world. The researchers used two days of computational time across several thousand GPUs. “I ran the code in parallel so that many different parameters, iterations of the same system and different proteins could be distributed across the GPUs,” explained Jens Glaser, a computational scientist at Oak Ridge, in an interview with ORNL’s Rachel McDowell. “This allowed us to easily make use of Summit’s parallel computing capabilities.”
The simulations bore out the strong role of shape in protein binding – at least, in some of the cases: with the simulations only accounting for the proteins’ shapes, six of the 46 protein pairs assembled more than half the time – and one of the pairs bound more than 94 percent of the time. “We’ve demonstrated that something as simple as shape is able to predict protein interactions that are sometimes really complex,” Glaser said. “This first demonstration has led us to believe that shape has been an unappreciated ingredient in many protein assembly processes.”
“Before we did this study, I actually didn’t expect proteins could form dimers based on shape alone,” added Fengyi Gao, a PhD candidate at the University of Michigan. “But now, we’ve found that this works, and we can study more complex structures or even combine this with other approaches, like machine learning, to see which features we need to enable the correct binding.”
Indeed, the researchers successfully built a machine learning model to detect proteins that could assemble based on shape alone, aiming to isolate the additional variables necessary to predict the binding of even more proteins. The team is also working on scaling the predictive approach up, and closer to real-world applications. “We think we can adapt this approach to something like drug screening in the future,” Gao said. “In addition to that, we hope that this shape-based model can serve as a basis of studying protein assembly in general.”