UW Bioengineering PhD alumnus Mike Kellen of Sage Bionetworks
Sage Bionetworks is a non-profit working to create open systems to hasten the pace of biomedical research. As director of technology platform and services, Kellen leads two teams: one that enables large-scale data sharing among life sciences researchers, and another that devises new ways to capture medical research data using cell phone and sensor platforms.
A foundation in software
As a bioengineering graduate student at UW in the lab of Professor James Bassingthwaighte, Kellen developed physiological modeling and simulation software that is still used in the open source community. Seattle-based software company Teranode later licensed and commercialized a version of his technology. After earning his Ph.D. in bioengineering in 2002 with a focus in computational biology, Kellen joined Teranode as the company’s first employee. He held positions in product development, field consulting, and development management until 2010, when he joined Sage Bionetworks.
New ways to collaborate on big data
Achieving an in-depth understanding of human diseases requires many different types of data from many sources, and a large number of clinical samples. “The notion that one small group or even one large company can lock up an area, understand it completely and get a product out is increasingly breaking down in life sciences and in the pharmaceutical industry, in which I interact a lot,” Kellen says.
Sage Bionetworks’ approach is to facilitate partnerships across institutions. Sage helps researchers join forces to tackle a large project and complex datasets. It also builds the general strategies and tools that can be used by many groups to solve diverse problems.
Building an open source culture
To encourage groups of people to choose to work together in a large-scale way, rather than work in their own individual lab groups, Kellen and his team developed Synapse, a platform that supports open, large-scale analysis of clinical genomics data. The technology may be powerful, he says, but first you need “the will to work together.”
Motivating people to want a collaboration to succeed in the first place and building trust is the greatest challenge, Kellen says. This can be even more difficult with complex technology and science problems. “Once they turn into technical challenges and not people challenges, it’s kind of inevitable that people find ways to make it work.”
Kellen says the vision for Synapse is a system in which scientists collaboratively build predictive models of human disease using large-scale, pooled data. Ultimately, these models could help evaluate a patient with a complicated disease such as cancer, Parkinson’s or Alzheimer’s –diseases that currently have huge variations in side effects and treatment effectiveness in different people –and then determine the best treatment options for an individual.
“I think the bottleneck is understanding the data, not generating the data,” Kellen says. “The measurement technologies are getting really good and really cheap amazingly quickly right now. The ability to use that data is lagging behind.”
A case study: colon cancer researchers pool resources
Over the last few years a number of groups conducted clinical and genomic studies on different patients with colon cancer. A string of papers from well-known groups followed, each explaining how they analyzed their data and found major sub-classifications of the disease, which can help guide treatment decisions. The first four teams published their articles, but none of them agreed on how the disease was breaking down into major sub-areas. By the time the fifth and sixth research teams published, there was massive confusion in the field about how the data was analyzed and what it meant, Kellen says.
Sage offered their Synapse technology platform to the group leaders, allowing them to pool their data. “We were able to get them to agree that if they each tried to do their own analysis on the pooled data, then they all had a reason to share their data with each other first and then eventually put that into the public domain and let other people work on it as well,” Kellen says. He expects the consortium of groups will release a paper soon that consolidates the independent studies and provides more consensus on the sub-typing classification system for the disease.
DREAM challenges, the ‘XPrizes’for bioinformatics
The Synapse platform also supports Sage’s DREAM Challenges, which are open competitions to crowd-source solutions to specific bioscience problems. In five recent competitions that finished in September 2014, more than 1,700 people signed up to compete. Challenge problems included identifying the best biomarkers for early Alzheimer’s-related cognitive decline and predicting how acute myeloid leukemia patients will fare after treatment. Four new challenges are currently underway – including projects to predict how a molecule will smell and forecast survival for prostate cancer patients. Prizes include invitations to speak at international conferences, publication in prestigious journals and high-profile publicity for the winners.
“It’s really interesting what happens when you don’t go to the five experts in an area and see what they think – you just leave it wide open,” Kellen says. “It is another aspect of leveraging data and the power of open communities of people in order to find solutions to problems.”
A bridge to social and sensor technology