the forefront of health and data science, is that we don’t know how to make those data sets talk to each other, and we don’t have the tools to analyze them.
At a public forum Friday in San Francisco dedicated to precision medicine, one of those experts, Atul Butte, noted that you can download 14,000 patient genotypes [not genomes, as reported in a previous version of this story] from the Framingham Heart Study, a nearly 70-year-old ongoing health study of the people in a town outside Boston. And that a U.S. government database called PubChem has a wealth of data from drug experiments that remains to some extent untapped. These and other data are waiting for researchers to build better analysis tools. Half the researchers in his lab are working on data related to populations, Butte told the forum Friday. [This quote from Butte has been changed to clarify its meaning.]
(Butte just moved from Stanford University to the University of California, San Francisco to start a new Institute of Computational Health Sciences. One of his goals is to tie all five UC medical centers, whose patients represent 4 percent of the U.S. population, into one data system.)
In other words, the bigger the sample sizes, the more accurately drug makers can design medicines for specific people, or the more confident health providers will feel recommending lifestyle changes and preventive measures for patients with particular disease risk factors.
That’s why the U.K. has a national “100,000 Genomes” project, and the Obama administration wants to build a health database with data from 1 million Americans.
On Monday, the National Institutes of Health named the members of the working group steering the project. The co-chairs are Richard Lifton from the Yale University School of Medicine; Bray Patrick-Lake from Duke University; and Kathy Hudson, NIH deputy director of science, outreach, and policy. (Among the rest of the panelists named Monday are Gates Foundation CEO Susan Desmond-Hellman, Yumanity Therapeutics CEO Tony Coles, and Multiple Myeloma Research Foundation founder Kathy Giusti.)
At the forum Friday, the White House’s precision medicine coordinator, Jo Handelsman of the Office of Science and Technology Policy, acknowledged the challenge of building the national database, or “cohort,” from a patchwork of smaller cohorts. Few of them have records that “talk” to each other, for example.
“The biggest challenge is the interoperability of electronic health records,” said Handelsman, who joined the meeting from the White House via Skype—after a few minutes of technical difficulties. “There are major barriers at the infrastructure levels.”
When asked if the government should force software makers to adopt uniform standards, Handelsman said the White House prefers to “coax them without demanding anything,” but hasn’t made any decisions. (Former FDA commissioner Andy Von Eschenbach, speaking an hour later at the forum, said, “I don’t think we can do it democratically. We should just do it, even if it means the government stepping in and telling industry, ‘This is the way to do it.'”) [A previous version of this story misspelled Von Eschenbach’s name. We regret the error.]
What about the U.S. health database tapping into stores of private information? 23andMe has compiled enough data to start its own drug-hunting division, Google has begun a collection of health data from healthy people called “Baseline,” and Apple is letting users of its new wrist watch funnel personal data to health studies via software it calls ResearchKit.
Handelsman praised those new efforts, but she cited security problems and patient biases in commercial data. (One worry, for example, is that the Apple-centric data would overly represent educated, well-to-do people because that’s who buys top-line Apple products.)
But to be fair she noted that no single database will hit all or even a majority of the criteria the U.S. project is shooting for—all the ‘omics, plus diet, exercise, and lifestyle data—or has the necessary diversity. A big question is how many of the 1 million Americans the national database wants to include will have to be recruited afresh to fill the gaps left after stitching all the outside databases together.
Whether the benefit of all this detailed information remains mainly the provenance of smaller projects, like deCODE’s in Iceland, or eventually bears fruit through vast databases, many advocates are adamant that success will only come through a radical shift in the patient’s relationship to the healthcare system. Right now that relationship is “feudal,” argued Stephen Friend during Friday’s forum. Patients are “subjects” poked, prodded and mined for their data. In fact, said Friend, we should stop using the word “patient,” preferring instead “participant.”
A former Merck research executive, Friend came back to Seattle in 2009 to launch Sage Bionetworks, a nonprofit dedicated to fostering open-source collaboration in biomedical research. (I wrote about its Alzheimer’s project as part of this feature last summer.)
During the forum, Friend said a “participant-centered model” will be a positive disruption, and pointed to Apple’s ResearchKit sharing system—the user gets a clear, simple consent option before any health data are transmitted—as a good start. (Sage has released two research apps, one for Parkinson’s disease, another for breast cancer recovery. Here is the consent form specific to the breast cancer app.)
Researchers like Friend have little choice. To get the data needed to build big, complex pictures that yield pinpoint solutions, people—patients, participants—will have to trust the collectors. Icelanders, to some extent, already do, as do the surprising number of people who signed up immediately to share their data via ResearchKit. The U.S. government can only hope Americans are just as enthusiastic about the national project.
