healthHealth and Medicine

IBM Watson set to battle brain cancers


Janet Fang

Staff Writer

clockMar 21 2014, 04:49 UTC
495 IBM Watson set to battle brain cancers
A cancer mutation is shown on a cell protein pathway from genome sequencing / IBM Research
The age of personalized medicine promises treatments tailored to you, based upon extensive analysis of your DNA to determine how a given medication would affect you. The problem is, that takes time -- a cadre of doctors are required to analyze your genetic makeup and craft a treatment specific. However, there's a new tool that could accelerate the development of future treatments: IBM's Jeopardy!-playing supercomputer, Watson.
Since its notorious stint as a game show champion in 2011, Watson has been training for a new career in medicine. The team at IBM thought Watson's talents for quickly scanning enormous data sets and understanding natural-language questions could make the artificial intelligence adept at diagnosing illness. Now Dr. Watson is taking on an aggressive cancer as a test case for its medical prowess.
The Watson team is focusing on a deadly brain cancer called glioblastoma. Patients have a median survival time of barely more than a year -- not good when the best treatment involves time-consuming research, studying samples of cancerous and non-cancerous cells in the patient to find a way to target the cancerous ones. Now Watson is on the case. As part of a new study with the New York Genome Center, the IBM DeepQA system will analyze the genetics of 20 to 25 glioblastoma patients to try to filter out something useful from the mountain of data. It'll then compare that data to the patient's clinical information so it could make treatment recommendations.
Compared to even an expert human, Watson can sort through this genetic data deluge in a heartbeat. It can sequence 75 million base pairs of DNA in a second, and that's good, because studying one person's brain tumor might require sequencing 800 billion pairs. Each person's tumor is genetically unique, but there could be thousands of mutations that created differences between a normal and a cancerous cell. Watson has to find those differences, sort out which ones are the most important to the development of this disease, and see what kinds of treatments worked in the past against cancer cells that were similar to those in its current patient.
So speed helps, but it's not the only thing. IBM says Watson is such a good tool for medical research because it learns. Programmers have already spent years teaching Watson to understand the language of medicine and how parts are connected in the body and brain. As Watson scans more and more data and research papers, it gets "smarter," and can recognize patterns faster. And the IBM team is starting to include into Watson's decision-making calculus the knowledge of which of its own treatment suggestions succeeded and failed in past cases.
Watson won't replace your doctor anytime soon -- humans still make the final calls about treatment. But with IBM's help, medical advice that otherwise would require a team of specialists and a week's worth of time can be had almost instantly. 
Image: IBM Research via Flickr

healthHealth and Medicine
  • tag
  • cancer,

  • watson