Topple the Walls, Open the Data

Matt Luchette 

 

May 7, 2014 | If Stephen Friend shares anything with Ronald Reagan, it’s his solution to complex problems: tear down this wall.

 

Friend, the president of Sage Bionetworks, gave the keynote presentation at the Bio-IT World Conference & Expo Wednesday morning on enhancing communication within the biomedical research community. Friend’s communication concerns span every level of research, from creating protein pathway models, to analyzing patient data.

 

Complexity is Key 

 

Biology has become too much like storytelling, said Friend, and searching for a narrative in our data has hidden the complexity of biological systems. 

 

“We like story-telling, […] but that’s not how we got through evolution,” he explained. 

 

The problem, he said, stems from grant agencies like the NIH who prioritize hypothesis-driven experiments, causing researchers to “extrapolate from the particular to the general.” In trying to build a narrative, researchers draw more sweeping conclusions than their data may suggest. The technique is fantastic for headlines (“Gene Predicts Time Of Death Down To Hour, Study Suggests”), but it creates a dogma that’s difficult to upend. Researchers analyzing a RAS pathway for potential therapeutics may each have different results, but as Friend explained, “that’s not noise, that’s signal.” Instead, the research community needs to learn how to embrace that complexity.

 

Make Them Compete 

 

Embracing complexity requires experts, especially in multidisciplinary fields like cancer, and as Friend explained, “The person who generated that data may not be the best person to analyze it.” The solution, he says, is to move away from biology’s publication model for sharing data—which can delay a study’s published results for a year after the study was completed. 

Biomedical research could learn from the strong community in software development: Web-based services like GitHub, allow software engineers to collaborate on projects in real time. “We should have a GitHub,” he suggested. “I’m jealous of the software engineers.”

 

But this would discard one of the key features of the publication model: attribution. The key, says Friend, is building into these sharing tools ways for scientists to get credit for their work. “Prominence is essential,” he explained.

 

Synpase, a Web service from Sage Bionetworks, is Friend’s attempt to bring GitHub’s spirit to the lab bench. The idea is simple: big data has given us more genomics and clinical data, but the same number of scientists is left to analyze it. Tools that streamline communication can let more experts analyze a group’s data. Errors are caught faster, and a more diverse group of experts can get involved with a project. 

 

By hosting the TCGA data (The Cancer Genome Atlas) for cancer genotypes on Synpase, Friend said, the program allowed over 100 researchers from 30 institutions to work with the data. In nine months, the groups were able to publish 12 papers on the data.

 

Despite his enthusiasm, Friend is hesitant about how seamlessly such a program would be adopted. “I’m much more worried about the academics” than those in biotech, he said. The biotech industry has learned how to work in teams, but the tenure model of academia has made professors more reticent. 

 

One solution, he suggests, is competitions. The DREAM Challenge, hosted by Synapse, asks participants to use Alzheimer’s patients’ genome sequencing data to build a model that predicts cognitive changes over time. “If you want people to share,” said Sage, “let them compete.” 

 

Patients on the Team 

 

Even ensuring efficient communication between researchers leaves out a critical variable: the patients they hope to treat. “We’re doing a good job of getting genotypic data,” said Friend, “but we’re not doing a good job of getting phenotypic data.” 

Companies like 23andMe aggregate and monetize patient data by creating an “entertainment” service, he says, which lets clients learn about their genes, while contributing to the company’s patient data bank. But Friend thinks research can benefit by treating patients more like research partners. 

 

“We’re interested in turning anecdotes into signal,” said Friend, echoing Sage’s motto. Sage is developing a mobile app for Parkinson’s patients to upload their subjective data to help researchers learn more about who’s benefiting from treatment.

But on all levels, from pathways to patients, biological research needs to open its gates, said Friend. Instead of secrecy and scoop scares, he insisted, the researcher’s instinct should be, “How can I get the most number of people looking at my data as possible.” 

 

“What’s it going to take,” he asked, “to have that be the way we think?”