Heart arrhythmia can prove deadly, contributing to the hundreds of thousands of deaths from cardiac arrest in the U.S. every year. Unfortunately, many of those arrhythmia are induced as side effects from various medications. Now, researchers from the UC Davis School of Medicine have leveraged supercomputing resources allocated by the Extreme Science and Engineering Discovery Environment (XSEDE) to work to better understand which medications are likely to induce those troublesome arrhythmia. 

Arrhythmia-related side effects have been a leading cause of medications being removed from the market for decades, leading to an FDA-mandated procedure during drug testing that measures the average time between two waves (“Q” and “T”) on an electrocardiogram in patients – the more the wave elongated when the medication was taken, the higher the risk of arrhythmia. However, false positives were a problem – grapefruit juice, for instance, could do the same thing – which weakened the predictive power of the QT interval and potentially led to unnecessary rejections of otherwise-helpful drugs.

So, at the UC Davis School of Medicine, researchers decided to tackle this gray area with supercomputing.

“What we set out to do was to try to solve that problem by building a computer-based pipeline for screening,” said Colleen Clancy (a professor in the Department of Physiology and Membrane Biology and the Department of Pharmacology at UC Davis) in an interview. Clancy and her colleagues chose two drugs that both prolonged the QT interval: dofetilide, which is known to cause arrhythmia, and moxifloxacin, which is known to be very safe for most people. 

The researchers ran multi-scale simulations of the drugs’ interactions with a critical potassium channel in the heart, culminating in an all-atom simulation of the interaction lasting several microseconds. They then fed the results into a machine learning model to predict the cardiotoxicity of each drug. Using this pipeline, they were able to show the distinctions in how the two drugs affect arrhythmic risk.

“The big challenge computationally is the system that we studied is pretty large,” said co-author Igor Vorobyov, an assistant professor in the same departments as Clancy. “It’s on the atomistic scale. We have around 130,000 atoms in our system. […] Here is where supercomputers come in very handy.”

To run the intensive simulations, the researchers were awarded XSEDE allocations on Stampede2 at the Texas Advanced Computing Center (TACC) and Comet at the San Diego Supercomputer Center (SDSC). They also used time on Blue Waters at the National Center for Supercomputing Applications (NCSA) and Anton2 at the Pittsburgh Supercomputing Center (PSC). 

“Stampede 2 offered a large array of powerful multi-core CPU nodes, which we were able to efficiently use for dozens of molecular dynamics runs we had to do in parallel,” Vorobyov said. “Such efficiency and scalability rivaled and even exceeded other resources we used for those simulations including even GPU-equipped nodes.”

“Now, we are not limited by our own resources,” he added. “We can use the top computational resources in the world to do these calculations. It totally changes your perspective as a scientist that you can use these resources to advance science, and feel like you belong to this large community of other scientists, for the greater good of the U.S. and the worldwide community.”

To read the article by TACC’s Jorge Salazar reporting on this research, click here. The study, “A computational pipeline to predict cardiotoxicity,” was published in the April 2020 issue of Circulation Research and is available here.