Schnieders' Need for Speed and GPUs

As anyone suffering from an illness knows, time is physically and psychologically their worst enemy. For the medical and insurance industries, time = money.

UI Biomedical Engineer Michael J. Schnieders’ research lies at the intersection of biochemistry and engineering, with an emphasis on molecular biophysics theory and high-performance computing (HPC) algorithms needed to reduce the time and cost of engineering drugs and organic biomaterials. His work takes advantage of the 100 Graphics Processing Units (GPUs) on UI’s Argon supercomputer to accelerate the process of discovery.

Schnieders holds a D.Sc. in Biomedical Engineering from Washington University in St. Louis, Missouri and completed postdoctoral fellowships in Chemistry at Stanford and in Biomedical Engineering at the University of Texas at Austin. “During my Stanford and UT postdocs, I was exposed to a range of supercomputers including GPU computing on Folding@Home and Intel coprocessors at the Texas Advanced Computing Center (TACC). The ability to continue to use state-of-the-art computing hardware at the University of Iowa via the Argon cluster is critical to our research program.”

The relationship between computational science and medicine has evolved to where patient-care decisions and drug design are increasingly reliant upon advanced computation and big data. Since the introduction of GPUs in 2010, the time to solution for many fluid and molecular dynamics challenges became 10 to 20 times faster than with traditional central processing units (CPUs) making it possible to sequence the human genome for around $1,000, where ten years ago it cost $100 million. This advent launched the science of genomics and the frontier of precision medicine.

With the ability to classify a broader range of molecular phenotypes, predictive analyses and treatments can be more accurate, and personalized care is affordable to a broader population. Non-invasive, virtual analyses that are informed by patients’ genetic makeup lead to safer and earlier detection, and more effective treatment. There is less trial-and-error involved with the time-consuming task of understanding patient histories and choosing from what can often be a broad range of prospective treatments.

In the past, especially for cancers and neurological diseases, like Alzheimer’s, Parkinson’s and Amyotrophic Lateral Sclerosis (ALS, or Lou Gehrig’s Disease), a diagnosis and course of treatment couldn’t be offered without multiple visits to a series of specialists over a period of months—even years, in some cases. Now these steps can often be resolved in a matter of minutes on the first visit by a computationally-savvy diagnostician who has access to the right instrumentation and data. Medications are also “smarter” as they are designed to dissolve at a specific temperature and release just the right medicine at intervals without over- or under-dosing the patient for the quickest and most efficient results.

Schnieders’ team is exploring solutions to all of these problems and others that are associated with an aging population. For example, hearing loss affects every octogenarian and one in 500 newborns. His team is contributing protein structures to the Deafness Variation Database ( that categorizes deafness by a wide range of genetic traits and phenotypes, for example causation due to exposure to noise or aging. This will ultimately reduce the time required to identify a genetic predisposition to hearing loss so that precautions can be taken to prevent exposure to conditions that accelerate deafness.

“It once took an entire year on 20 compute nodes to optimize all of the protein structures included in the Deafness Variation Database for analysis of specific phenome-genome combinations that led to hereditary hearing loss; these calculations only require a week on 10 GPUs,” said Schnieders.

“When optimizing legacy code to take advantage of the acceleration GPUs offer, finding the sweet spot can be challenging, at first, but it’s well worth it,” he added based on experience with their computer code called Force Field X. With a range of computationally-intensive research that benefits from GPUs—including the many geospatially-implicit applications for edge computing and sensors—the UI Research Services and Informatics support teams are experienced with the code optimization process and can help researchers get the most out of Argon’s CPUs and accelerators. 

For more information about computational resources and services available to the UI research community, visit the ITS Research Services website.