Can a computer select the best cancer treatment?
Watson moves from Jeopardy win to doing doctors’ legwork
After IBM computer technology proved it was smarter than its champion human competitors in an epic $1-million Jeopardy tournament in 2011, scientists wondered: Might artificial intelligence do a better job than even highly trained physicians at predicting which drugs are best for cancer patients?
Scientists in Vancouver have now joined the North American search for that answer.
Researchers at the B.C. Cancer Agency’s Michael Smith Genome Sciences Centre (GSC) are the first and only scientists in Canada to get involved in a major research collaboration exploring how useful the technology can be when it comes to choosing therapies based on the DNA of a tumour.
The GSC has teamed up with just over a dozen top cancer centres in the U.S., all of them collaborating with the IBM subsidiary Watson Health.
We’ll compare what the computer did to see if it missed something. I expect Watson will need tweaking because we have to train it.
STEVEN JONES ASSOCIATE DIRECTOR, GENOME SCIENCES CENTRE
Watson is the name given to the Jeopardy champ — the “computer brain” turned into a cancer treatment prognitician. It’s named after the first CEO of IBM, Thomas J. Watson, and is billed as a bot that can do a lot of legwork for physicians, “reading” or mining databases of cancer gene analytics and advanced algorithms in a cloud-based system. Some of its sources include published treatment guidelines, research studies and a deep well of medical literature.
The goal is to find out whether Watson should be consulted for near-instantaneous searches on personalized treatments for cancer patients, especially involving those whose cancer has spread.
Steven Jones, associate director and head of bioinformatics at GSC, says the molecular biology of many cancers are so complex that doctors could use some technological help “juggling” all the cutting-edge information.
“People have been applying machine-based learning to biological problems for a very long time now. Here, we’ve got a large computational group and a machine that won Jeopardy,” he said, referring to the excitement that scientists like him are feeling about Watson’s potential.
Jones said in the first phase of the research, scientists will compare the treatment doctors chose to give patients against the treatments Watson selects as the most efficacious. Watson may be far from its origins as a bot loaded with trivia, but it is still in an evolutionary stage, so it won’t be used just yet to change doctors’ minds until its recommendations have been validated and peer reviewed. IBM says it works like this: “Watson completes the genetic material and medical literature review process in only a few minutes, producing a report and data visualization of the patient’s case, and evidencebased insights on potential drugs that may be relevant to an individual patient’s unique DNA profile.”
Jones says only some of the patients who have had gene sequencing on their cancer through the agency’s Personalized Onco-Genomics (POG) program will have their data entered into the study. The only patients qualifying for that program are those whose cancer is considered uncontrolled.
“We’ll compare what the computer did to see if it missed something,” says Jones, noting that Watson is expected to become smarter over time through feedback from researchers and any corrected mistakes. “I expect Watson will need some tweaking because we have to train it.”
So should doctors be worried that computer system software might one day replace them, especially since robotic technology has already destroyed the need for so many jobs?
Jones says while personalized medicine using gene sequencing is the “future of oncology,” he thinks physicians need not worry; the DNA sequencing field has always been heavily reliant on computer data generation and analysis. It’s not like patients will be getting their prescriptions from a robot replacing oncologists; the computerized recommendations or responses from Watson — or an entity like it — would merely be passed on to doctors as an adjunctive decision-making tool that can accelerate the process.
IBM describes the research as part of its broader goal of using Watson to “advance patient-centred care and improve health.
“As participating institutions use Watson to assist clinicians in identifying cancer-causing mutations, Watson’s rationale and insights will continually improve, providing the latest combined wisdom of the world’s leading cancer institutes for oncologists.”
Watson doesn’t provide absolute answers. It offers possibilities with confidence ratings to let doctors know whether the treatments are merely experimental, the gold standard or something in between.
Jones said IBM representatives visited Vancouver and were duly impressed with how advanced the gene-sequencing facility is, and to see the POG program in place.
The collaboration didn’t cost anything to join and there’s no closing date on the research so “we’re just going to keep going until we have a body of knowledge to publish.”