Sign Up for Fundamentals

Stay up-to-date with the latest research findings from the Institute for Basic Biomedical Sciences.

Please enter a valid email address.
Fundamentals Topics+

Creating a Big Data Platform for Personalized Health

More Advancements in Research

Creating a Big Data Platform for Personalized Health

Creating a Big Data Platform for Personalized Health

By Vanessa Wasta

February 2018 — The zoomed-in view from Google Earth on a set of buildings in Northern Virginia is nothing spectacular. But the nondescript, rectangular buildings hide the impressive facility within. It’s built to withstand the most severe natural disasters, and no one gets in or out without passing strict security measures. Its power generators could fuel a small city.

Behind this curtain of concrete and steel is an epicenter of computer data storage and capacity. Such “data centers” have become the go-to facilities for many companies with complex computing and data needs. One area with the most explosive growth in data storage needs is medical research.

With huge troves of genomic and other types of research data, the so-called “under-the-desk” computer server is vastly underpowered to handle such large data sets. Some 30 faculty members at Johns Hopkins Medicine are now working with staff members in information technology to host research data on Microsoft- and Amazon-based cloud computing systems.

Steve Sears, director of cloud services, says it’s cost effective to host data on the cloud. “We pass through a discount, based on economies of scale. There are no desktop charges, and we only charge for specific configurations,” he says.

The potential for cheaper, more efficient ways to handle big data has implications for precision medicine efforts as well.

In precision medicine initiatives, data come from many sources, such as research, radiology, medical records, and wearables. “We need to put all of this data in one place, and, at the same time, shrink the effort and cost of finding important data and building the infrastructure for it,” says Aalok Shah, director of product management.

Once considered the mecca of all medical data, digital health records have not panned out as a panacea for precision medicine. Shah says researchers realized that health records are best suited for day-to-day operations of hospitals and medical practices, not for housing research data or the massive files of radiology scans. Add in data from wearable technology, and the medical record seems less likely as a place where all of this information can converge easily.

Faculty and staff members at Johns Hopkins inHealth, which focuses on precision medicine initiatives, are working with scientists and analysts at the Johns Hopkins University Applied Physics Laboratory to develop a “precision medicine analytics platform.” PMAP, as it’s called, will funnel data from a variety of sources, including Hopkins’ electronic medical record system, into a cloud-based platform to help scientists connect treatments and outcomes with patterns of genomic, phenotypic and other patient-related data.

PMAP’s prototype is rooted in one of the most hotly debated fields of medicine: prostate cancer. Kenneth Pienta, M.D., director of research for the James Buchanan Brady Urological Institute, and his colleagues at the Johns Hopkins Technology Innovation Center, have been working for several years on an app for both clinical decision support and education for patients diagnosed with prostate cancer that is low-grade—meaning not aggressive—and who are at low risk for worsening disease.

testimg Ken Pienta speaks at the Precision Medicine World Conference in Silicon Valley. January, 2018.

These men are more likely to die of other causes than prostate cancer. The field has shifted from invasive and side-effect-prone prostate-removal surgeries to active surveillance programs that closely monitor prostate health. One of the largest such programs, which has amassed 20 years of patient data, is led by Johns Hopkins urologist H. Ballentine Carter, M.D. This data has led the field in defining which men are at low risk for disease progression and how to monitor their prostate health.

Leveraging the information from patients enrolled in the active surveillance program, Shah and Pienta are working on an app with Rebecca Yates Coley, Ph.D., a biostatistician and former postdoctoral fellow at Johns Hopkins who now works for the Kaiser Permanente Washington Health Research Institute.

At the recent Precision Medicine World Conference in California’s Silicon Valley, Coley asked, “What are the best methods for developing risk prediction models that provide individualized decision support in a clinical setting?”

Their app aims to do exactly that: Bridge the divide between discovery and patient care. “We want to help clinicians view patient data in a longitudinal way—for example, their patients’ PSA scores over time, the timelines of their condition, percentiles for PSA score—which can help give patients better context for their prostate cancer,” says Shah. The app will also use algorithms based on Johns Hopkins data to define factors such as an individual’s risk for an increase in cancer grade after a biopsy.  

“This helps our patients understand the factors that should go into the decision of whether to remain in an active surveillance program or undergo prostate surgery,” says Shah.

The app is not yet ready for wide use among urologists and their patients, says Shah. Hopkins faculty and staff members are beginning to test it now.

Considering wearable technology and such apps, Dwight Raum, vice president and chief technology officer for the Johns Hopkins Health System and The Johns Hopkins University, says there is an underlying transition in the way we are collecting patient data.

“There is a shift from a linear process of discovery, followed by applying that knowledge to the clinic, to analyzing real-time feedback and incremental changes within certain populations,” says Raum.

But, he says, we have to get researchers with a traditional mind set to think about how to expand data beyond increasing sample size. In addition, we must determine how to obtain and analyze data within the current framework of our institutional review boards for research, says Raum.

“The institute that cracks that nut will have a leg up on research,” he says.

Raum says this is where team-based discovery can be most effective: “This type of research takes a team anchored by PIs and complemented by data scientists to move the field forward.”