Stanford's annual Big Data in Biomedicine began this morning with a "breathtaking set of talks," as described by Russ Altman, MD, PhD, a Stanford professor of bioengineering, genetics and of medicine.
The first panel focused on genomics, with the speakers presenting a dizzying forecast of a future where biomedical data is standardized and easily accessible to researchers, yet carefully guarded to protect privacy.
"How do we build this in a way that allows you to spend time working on your science, and not spend your time to worry about reinventing the plumbing?," asked David Glazer, director of engineering at Google and a speaker on the panel.
His team is hard at work ensuring the infrastructure of the Google Cloud Platform can withstand the rigorous demands of a slew of big data projects, including the Million Veteran Program and MSSNG, an effort to understand the genetics of autism.
For panelist Heidi Rehm, PhD, associate professor of pathology at Harvard Medical School and director of the Partners Laboratory for Molecular Medicine, a key hurdle is standardizing definitions and ensuring that supporting evidence is available for system users. For example, data developers should be able to demonstrate why a particular gene variant has been deemed benign, and what definition of "benign" they are using, she said.
Her team has developed a star system, which rates sources of data by their credibility, giving results submitted by expert panels more stars than data submitted by a single researcher.
Rehm also addressed the pros and cons of various models to share data. Rather than collecting it all centrally, she said she expects data will be shared through a small number of hubs, which each have the ability to connect with each other, similar to an airline trafficking model.
Individuals are not standing in the way of research advances, reported panelist Jill Hagenkord, MD, chief medical officer of the personal genetics company 23andMe. She said that of their 950,000 customers, nearly 80 percent have agreed to share their data for research. Participants are also eager to provide additional information when asked, Hagenkord said. It becomes almost a philanthropic effort, they feel grateful that someone is interested in their conditions, she said.
Individual, average-Joe participation is also integral to panel member Yaniv Erlich's work. Erlich, PhD, assistant professor of computer science at Columbia University, has tapped the wealth of Geni.com, an online ancestry website, to develop family trees linking thousands, even millions of people.
Using the profiles of 43 million people, his team found that genetics explain about 20 percent of longevity and a variety of other striking findings, such as that as recently as a century ago, married couples were, on average, fourth cousins. One of his colleagues also developed a video map (see fourth video on page) that shows the migration of humans from western Europe.
Panelist Christina Curtis, PhD, an assistant professor of oncology at Stanford, shared a glimpse into the insights possible by applying big data techniques to just one individual -- even just one tumor. Her team has developed a "Big Bang" theory of colon cancer development, which stresses the importance of the development of early variations in tumor growth and meshes with findings of surprising heterogeneity within a single tumor.
Curtis' findings demanded a new data framework simply to interpret a single tumor, offering a warning to clinicians about the limitations of single tumor samples.
Big Data in Biomedicine isn't just a gathering of folks in Li Ka Shing Center for Learning and Knowledge. All of the action is available remotely using live-streaming, an app can be found at crowd.cc/bdbm15, and conversation is happening on Twitter with the hashtag #bigdatamed.
Previously: Countdown to Big Data in Biomedicine: Leveraging big data technology to advance genomics, Using genetics to answer fundamental questions in biology, medicine and anthropology and The Big Bang model of human colon cancer
Photo of Jill Hagenkord by Saul Bromberger
