Oncologists Partner with Watson on Genomics

Genetic sequencing has become increasingly affordable and accessible for cancer patients, but the complexity of processing and analyzing the data—and the time that that takes—means that relatively few patients truly benefit from the effort. That may soon change under a new collaboration between cancer centers and IBM's Watson, a powerful cognitive computing program capable of reviewing reams of data and pinpointing potential treatment options within minutes.

“With Watson, oncologists can quickly identify the most likely driver alterations that are causing the cancer and which drugs they should be thinking about for that patient,” says Steve Harvey, vice president of IBM Watson Health. “Our system is doing in a few minutes what it would take days or weeks to do manually.”

Participants in the program, which is available to subscribers, upload patients' sequencing data into the Watson Health Cloud platform, where it is analyzed by Watson Genomic Analytics. The program searches for actionable mutations and reviews the most-current evidence-based guidelines, clinical trials, journal articles, and patient outcomes. It then produces a report for the oncologist, listing potential treatment options alongside supporting evidence.

So far, 14 cancer centers have joined the project and are helping to expand Watson's capabilities by uploading clinical data and providing feedback on the quality of the results. For example, the University of North Carolina (UNC) Lineberger Comprehensive Cancer Center in Chapel Hill contributed anonymized sequencing data and patient outcomes from 1,800 cases reviewed by its Molecular Pathology Tumor Board. The board meets weekly to consider the mutations found in patients' tumors and identify potential targeted treatment options.

“We wanted to find out if Watson, presented with the same data, would make the same call as our tumor board,” says Norman Sharpless, MD, the center's director. “Right now, Watson isn't as good as people at deciding which mutations are actionable in a tumor, but I think we can fix that within a year by refining the algorithms it uses to analyze information.”

Sharpless then plans to conduct a clinical trial in which oncologists would receive two sets of recommendations for a patient—one from the molecular tumor board and another from Watson. The goal is to determine how often those two lists agree and, if they differ, which one is better at helping clinicians make treatment decisions.

Training Watson to select potentially effective drugs is by far the most challenging aspect of the project, says Sharpless.

“Ingesting the medical literature and keeping abreast of the field is a very powerful advantage of Watson,” he says. “But picking a drug based on that information isn't easy. Some of the best drugs we have—such as Taxol and platinum—aren't easily identifiable and what makes patients respond to those agents isn't always clear.”

Tackling the problem will require training Watson on a very large set of clinically annotated genomic data, which does not currently exist at any single institution, he says. Several large private and public organizations are working toward that goal, including IBM, Flatiron Health, Google Genomics, and the American Society of Clinical Oncology, as well as major health insurers.

Watson doesn't replace humans, but it is helping clinicians to work more efficiently and to conduct genomic sequencing and analysis on a much larger scale, says Sharpless.

“Today, sequencing doesn't help most patients, but I'm very optimistic that with the help of cognitive computing systems like Watson, the percentage will go from about 5% today to 50% or higher,” he says. “We will get much better at using genomic information to pick therapies that are beneficial.”

For more news on cancer research, visit Cancer Discovery online at http://CDnews.aacrjournals.org.