Paging Dr. Watson, IBM's medical adviser for the future
- 29 August, 2014 07:39
IBM Watson can match possible treatments to genetic disorders
IBM continues to make the case for the nascent field of cognitive computing, showing off some Watson prototypes that could help speed scientific discovery in the medical field, by scanning large volumes of literature and data far more quickly then humans can, and suggesting possible leads.
The New York event was to promote an update to a cloud service, called Watson Discovery Advisor, that can help researchers find fresh connections within mass realms of data.
IBM Research developed Watson to compete with human contestants on the "Jeopardy" game show in 2011, using natural language processing and analytics, as well as many sources of structured and unstructured data, to formulate responses to the show's questions.
In the years since, the company has been working to commercialize the Watson technology, by identifying industries that could benefit from cognitive computing, such as health care, law enforcement and finance.
In the presentations on Thursday, IBM senior vice president Mike Rhodin pointed to how Watson could speed literature searches and analysis for a particular field, a process that is increasingly becoming too unwieldy for even the largest research teams to complete.
The medical field serves as a good testing ground for these capabilities.
Medical information doubles every three years and by 2020 will double every 73 days, Rhodin said. As a result, a doctor can no longer rely solely on their own medical training, even with voracious reading of the latest medical journals. "We are human, and there are limits to what we can learn," he said.
Watson can do the background work necessary to narrow down the information to a smaller set of possibilities, Rhodin suggested.
"Watson is not bound by volume or format," Rhodin said.
In one presentation, IBM showed how it helped the Baylor College of Medicine more quickly identify possible treatments for a protein linked to many cancers, called p53.
More than 70,000 scientific papers have been written about p53. "It is just impossible for a single researcher to read all that. Yet there could be very critical facts in that evidence they can't get access to," said Scott Spangler, an IBM Research manager overseeing analytics. He referred to this under-read material as the "dark literature problem" of the medical community.
The Watson service was able to comb through papers and suggest a number of other proteins that could control p53 activity. Using Watson's analysis, researchers found seven proteins within a few weeks that could lead to possible treatment. Typically, the medical community finds about one new protein-based treatment a year.
Another organization testing Watson has been Johnson & Johnson, which also put the cognitive computing service to work ingesting and analyzing scientific papers, this time to compare the effectiveness of various drugs and other treatments, looking for those with the best results and the fewest side-effects.
For example, for back pain, a quick scan of the medical literature turns up published results of over 3,000 trials, describing the effectiveness of 27 different treatments. No one doctor can read all of these documents, said Soledad Cepeda, Johnson & Johnson director of epidemiology.
Also discussed Thursday was work from the New York Genome Center which, in partnership with IBM, has been conducting a clinical trial examining if Watson can help individuals with glioblastoma, an aggressive form of brain cancer. Watson can examine the patient's DNA data and look for potential treatments in the scientific literature.
Deriving a genetic treatment for each individual patient would take too long if done by hand, explained Bob Darnell, president of New York Genome Center, in a video presentation at the event.
The individual's DNA might have thousands of mutations, of which only a hundred or so may be relevant for the treatment. "They will be different for each patient with cancer," Darnell said. The scientific literature must then be searched for any possible drugs that could treat these specific mutations. This must be done quickly, Darnell said, given that the average patient may have only a year to live after diagnosis.
"What we need IBM Watson to do is to go through literature the same way we do now in two weeks, and do that in two seconds," Darnell said.