Supercomputer Simulates Beating Heart, Helping Drive Research

Sequoia is bringing modeling and simulation and modeling to new peaks, timing in at 16.32 sustained petaflops (20 PF peak), enabling scientists to capture greater complexity in a shorter period. With this advanced capability, Lawrence Livermore US National Laboratory’s (LLNL; Livermore, CA, USA) scientists have been able to simulate the human heart down to the cellular level and use the resulting model to predict how the organ will respond to different drug compounds.

The simulations were made possible by an advanced modeling program, called Cardioid, which was designed by a group of scientists from LLNL and the IBM T. J. Watson Research Center (Yorktown Heights, NY, USA). The extremely scalable code simulates the electrophysiology of the heart. It functions by breaking down the heart into units: the smaller the unit, the more effective the model. Up to now, the optimal modeling programs could attain 0.2 mm in each direction. Cardioid can jump down to 0.1 mm. Where earlier, researchers could run the simulations for tens of heartbeats, Cardioid executing on Sequoia captures thousands of heartbeats.

Scientists are experiencing 300-fold speedups. It used to take 45 minutes to simulate just one beat, but now researchers can simulate an hour of heart activity--several thousand heartbeats--in seven hours. With the less sophisticated codes, it was impossible to model the heart’s response to a drug or perform an electrocardiogram trace for a particular heart disorder. That sort of testing entails longer run times, which was not possible before Cardioid.

The model could hypothetically evaluate a range of medications and devices such as pacemakers to examine their effect on the heart, creating an avenue for more effective and safer human testing. But it is particularly suited to examining arrhythmia, a disorder of the heart in which the organ does not pump blood efficiently. Arrhythmias can lead to congestive heart failure, an inability of the heart to supply sufficient blood flow to meet the needs of the body.

There are various types of medications that disrupt cardiac rhythms. Even those designed to prevent arrhythmias can be detrimental to some patients, and researchers do not yet completely understand precisely what causes these negative side effects. Cardioid will enable LLNL scientists to examine heart function as an antiarrhythmia drug enters the bloodstream. They will be able to identify when drug levels are highest and when they decline. “Observing the full range of effects produced by a particular drug takes many hours,” noted computational scientist Art Mirin of LLNL. “With Cardioid, heart simulations over this timeframe are now possible for the first time.”

The Livermore-IBM group is also working on a mechanical model that simulates the contraction of the heart and pumping of blood. The mechanical and electrical simulations will be allowed to interact with each other, adding more realism to the heart model.

The cardiac modeling research was performed during the system’s “shakedown period”--the set-up and assessment stage--and the scientists had to rush to complete in the allotted time span. Once Sequoia becomes classified, it is uncertain if it will still be available to run Cardioid and other unclassified programs, although access will certainly be more difficult since the computer’s key mission is running nuclear weapons codes.

Sequoia is an integral part of the US National Nuclear Security Administration’s (NNSA) Advanced Simulation and Computing (ASC) program, which is run by partner organizations LLNL, Los Alamos US National Laboratory, and Sandia National Laboratories. With 96 racks, 98,304 compute nodes, 1.6 million cores, and 1.6 petabytes of memory, Sequoia will help the NNSA fulfill its mission to “maintain and enhance the safety, security, reliability, and performance of the US nuclear weapons stockpile without nuclear testing.”

The Cardioid simulation has been chosen as a finalist in the 2012 Gordon Bell Prize competition, awarded each year to recognize supercomputing’s ultimate achievements. The scientists divulged their findings at the Supercomputing Conference in Salt Lake City (UT, USA), on November 13, 2012.

Related Links:
Lawrence Livermore U.S. National Laboratory
IBM T. J. Watson Research Center
US National Nuclear Security Administration