IBM Research and the New York Genome Center want to prove that IBM's Watson can dramatically reduce the time required to identify the most effective cancer treatments based on an analysis of a specific patient's genetic mutations.
In partnership with regional hospitals, the NYGC plans to evaluate Watson's ability to help oncologists develop more personalized care for patients with glioblastoma, an aggressive and malignant brain cancer that kills more than 13,000 people in the US each year. The NYGC is a nonprofit consortium of the state's academic, medical, and university leaders working to advance the applications of genomics, the branch of genetics that studies complete DNA sequences rather than individual genes.
Precision medicine techniques based on genomic analysis have particularly great potential for cancer, according to Ajay Royyuru, a computational biologist with IBM Research.
"The difference [between] one individual [and] the next for the same diagnosis of cancer is humungous -- to the extent where a certain treatment is right for you, or a certain treatment will do no good for you," Royyuru told us. "Why is that? Because cancer is a disease of the genome." Not only are the abnormalities causing cancer different from one individual to the next, but they also change over time, becoming more and more abnormal.
That makes it particularly important to do the analysis leading to a treatment recommendation as quickly as possible, before the genome changes again.
The collaboration between IBM and the NYGC announced this week uses a customized version of the Watson cognitive computing system, which first became famous for winning a 2011 Jeopardy match over the game show's top human champions. Since then, IBM has made healthcare a primary target industry for commercialization of Watson technology and has worked with other institutions on applying Watson to decision support for cancer treatment.
The New York Genome Center research collaboration will use a new Watson prototype that has been trained to understand genomics data, accepting an analysis of cancer mutations as an input. Using the same skills that allowed it to extract mountains of Jeopardy trivia from news stories and reference sources, Watson will search the scientific literature to match genetic abnormalities with drugs associated with treating those mutations.
Target: glioblastoma (Source: Wikimedia Commons)
Toby Bloom, NYGC deputy science director for informatics, told us the current process of doing those literature searches is relatively manual and can take weeks or months. "We can sequence genomes relatively quickly, but then there's this big lag with how you associate that with drugs." Watson promises to reduce that process of weeks or months to minutes or hours, she said. The reality of that claim, which is based on IBM's tests against reference data from the National Institutes of Health's Cancer Genome Atlas, remains to be proven through the clinical trials IBM and the NYGC are now planning.
The promise is not that Watson will produce magical cures or even identify the one perfect drug for every condition. But by speeding up the process of identifying a short list of the most promising drug candidates, the technology makes it more likely that the treatment can be delivered in time to make a difference.
"As a first pass, we're just making it possible to select drugs quickly enough that are targeted for specific patients," Bloom said. "Assuming can find any improvements in at least a small number of patients, we will extend this study." Also, if the project starts seeing positive results, it should be possible to feed those outcomes back into Watson, "so the dynamic learning part of Watson could possibly find associations we wouldn't find as quickly."
Download Health IT in the Obamacare Era, the InformationWeek Healthcare digital issue on the impact of new laws and regulations. Modern technology created the opportunity to restructure the healthcare industry around accountable care organizations, but IT priorities are also being driven by the shift.
David F. Carr oversees InformationWeek's coverage of government and healthcare IT. He previously led coverage of social business and education technologies and continues to contribute in those areas. He is the editor of Social Collaboration for Dummies (Wiley, Oct. 2013) and ... View Full Bio
InformationWeek Tech Digest, Nov. 10, 2014Just 30% of respondents to our new survey say their companies are very or extremely effective at identifying critical data and analyzing it to make decisions, down from 42% in 2013. What gives?