In an effort to break down the walls that have long existed between research and clinical care, Geisinger Health System and Regeneron Genetics Center (RGC) partnered for the DiscovEHR collaboration aimed at coupling high-throughput, whole-exome sequencing data with de-identified electronic health records (EHRs). The result of the project allows researchers to link genetic variants to clinical phenotypes and disease diagnoses in medical records, according to two paper published in the December 23, 2016 issue of Science detailing DiscovEHR.
“We came with a big vision to do a very large-scale sequencing project, and we were thrilled to find someone who was like-minded,” says Aris Baras, vice president and co-head of RGC, of RGC’s partnership with Geisinger. Baras anticipates the DiscovEHR project will have collected DNA sequencing data from 100,000 participants by mid-2017 with an ultimate goal of 250,000 in the next three to five years.
In contrast to similar projects, which probe a limited set of genes, DiscovEHR sequences the entire genome. This wealth of genetic information, in conjunction with EHR data, has built a framework for discovering previously unidentified associations between genetic variants and disease—an aspect of the study that, according to Baras, has “turned out to be wildly successful.”
Perhaps an even more unusual aspect, however, is that Geisinger returns clinically actionable results back to study participants, informs primary care providers, and appends the information to EHRs. “In the old days, research and clinical care had a firewall between them,” says David H. Ledbetter, Ph.D., Geisinger Health System executive vice president and CSO. “We’re at the forefront of a new ethical framework around medical research that says we have an obligation to give clinically-relevant information back to patients.”
This unusual practice highlights Geisinger’s goal to identify high-risk patients and determine if preventive measures, such as enhanced surveillance or medication, can decrease the occurrence of life-threatening diseases like cancer or heart disease. DiscovEHR’s analysis found that 3.5% of healthy adults possessed a variant that placed them at significantly higher risk for early cancer, heart attack, or cardiovascular disease. The prevalence of these genetic variations could support the case for population-level genetic screening—if preventive measures can improve the outcome.
“The key is that the results have to be actionable, reliable, and there has to be a real utility” for the adoption of any new disease screening strategy, underscores Baras. So while the advent of new sequencing technologies could make population-level screening feasible from a cost perspective, it won’t solve the equation. Despite guidelines assembled by the American College of Medical Genetics on gene-disease associations known to significantly increase risk, the interpretation of genetic data remains a significant barrier to adoption.
Other medical centers have embarked on projects similar to DiscovEHR. Dan Roden, M.D., professor of medicine and pharmacology, director of Oates Institute for Experimental Therapeutics, and principal investigator of BioVU at Vanderbilt University, used the BRCA1 gene to illustrate the barrier to adoption of genomic data. “Suppose somebody finds a BRCA1 variant and there’s no family history of cancer. How seriously should they take that variant?” Dr. Roden asks. “There is no question that there are BRCA1 variants that don’t confer risk. Genetics shades risk, it doesn’t make it black or white.”
Vanderbilt created its DNA Biobank, called BioVU, almost a decade ago. With a collection of approximately 230,000 samples, “it’s the largest DNA collection coupled to EHRs at a single academic institution anywhere in the world,” Roden states. Although focused on research and discovery, the initial impetus for BioVU came, not from a desire to identify new genetic associations, but rather from a desire to act on known ones. “I was walking to work on a beautiful spring day, and I thought to myself, ‘you, and the field have defined many pharmacogenetics variants important for determining drug response, and yet nobody uses them,’” he recalls.
It occurred to him that preemptively embedding personal genetic information into a patient’s EHR could provide a realistic, if not futuristic, avenue for implementing the pharmacogenetics data he and his colleagues had collected. Vanderbilt’s PREDICT (Pharmacogenomic Resource for Enhanced Decisions in Care & Treatment) program has put this concept into practice by proactively embedding pharmacogenetics data into the EHRs of patients receiving care at Vanderbilt University Medical Center.
Both Geisinger and Vanderbilt are pushing the envelope used to predict disease risk or drug response, the future of personalized medicine will rely heavily on both the ability to interpret genetic data and to implement those conclusions in a clinical setting. EHRs will play an essential role in addressing questions of both interpretation and implementation, because, as Baras insightfully surmised, “this is not the future of medicine—it is the present of medicine.”
To see more articles from the January/February issue of Cliniclal OMICs click here.
