The problem is: This approach doesn’t scale. Today, figuring out how genetic mutations may impact a patient's response to a treatment is a time-consuming and manual process. It often involves a team of a half-dozen or more experts combing through hundreds of journal articles and consulting with colleagues around the world for a month. Even then, there's no guarantee that this will lead to useful information for the patient.
With more than 1.69 million Americans predicted to be diagnosed with cancer this year, there will likely only be a lucky few whose oncologists have the expertise, time and access to do this. The rest will be treated by doctors who have only as little as 15 minutes to make their treatment decisions.
IBM and the New York Genome Center are hoping to change this by partnering in a project that uses Watson, the computer superbrain of "Jeopardy!" fame.
As part of the White House Precision Medicine Initiative, on Thursday the organizations announced the creation of a national cancer tumor repository that they hope will pinpoint in a very broad way which genetic characteristics correspond to how people might react to the universe of all the treatments available.
"Our goal is to take that massive data and turn it into information that can be used for a patient who is waiting at the bedside," Robert Darnell, chief executive and founding director of the New York Genome Center, said in an interview.
The initial project will begin with 200 patients being treated at New York City-area hospitals, such as Memorial Sloan Kettering and Columbia-Presbyterian. Recruitment of patients begins immediately and is focusing on the neediest patients who have not responded to multiple treatments.
Unlike smaller precision medicine projects that focus on 30 to 50 genes, the IBM-New York Genome Center effort will involve whole genome sequencing of the estimated 22,000 genes in the body and look at many different aspects of DNA.
Steve Harvey, vice president for IBM Watson Health, explained that one way to understand the importance of the difference is by thinking about the iPhone camera, which typically generates a 3 megapixel image that is 2,000 by 1,500 little dots or 3 million bits of information. If that was the whole genome, he said, the genetic sequencing typically done at hospitals and direct-to-consumer companies only generates 3,000 of those dots.
"The resolution would be almost uninterpretable," Harvey said. "To really understand cancer, you would need to have the 3 million dots, which is what we are trying to do with the human body."
One aspect of the analysis will involve comparing DNA from tumor cells with DNA from other parts of the body. Darnell said the DNA in the different cells in a person's body should be almost identical except for the parts that are "broken" in cancer. The amount of data generated and analyzed for each patient could be more than 1 terabyte (1 million million bytes) — roughly equivalent to 2 million average-sized photo files.
The researchers will use Watson to spot patterns in that data and find links between that information and the more than tens of thousands of medical journal articles that have been published about cancer in order to discover which genes are most important in response to cancer treatments
Darnell and Harvey said that they eventually hope to scale up the project so that it can handle upward of 100,000 patients a year and that the repository will be accessible to researchers whose patients are participating in the project.
For more health news, sign up for our weekly newsletter.