In 2017, we’ve reported on everything from a cancer “kill switch” and AI-assisted cancer detection, to the first tests of a cancer vaccine and the creation of a portable skin cancer detector. These stories were just the tip of the iceberg in terms of exciting developments in cancer research this year.
While most of us would find it a little daunting to try to stay on top of the latest research, every piece of research is valuable. While some papers strike a chord with the media and receive a lot of attention, there are other papers and datasets that are inevitably missed, or deemed “uninteresting” or “irrelevant” to the general public at first pass.
A team of researchers from the University of North Carolina Lineberger Comprehensive Cancer Center set out to solve this conundrum, demonstrating their findings back in November.
Their solution involves using cognitive computing to sift through the overwhelming amount of data produced by scientific studies and databases in an attempt to “identify potentially relevant clinical trials or therapeutic options for cancer patients based on the genetics of their tumors.” Put simply, they want to use advanced computing techniques to more effectively identify potentially useful cancer treatments.
The researchers’ findings, published in the journal The Oncologist, suggest this new method could help physicians stay on top of current and upcoming scientific literature. To put the method to the test, the researchers enlisted the help of IBM Watson for Genomics to see if it might prove more effective than a panel of cancer experts. The team compared Watson’s ability to pick out treatments related to clinically significant genetic mutations with the findings of the board of cancer experts.
Of the 1,018 cancer cases they analyzed, the molecular tumor board identified actionable genetic alterations in 703 cases. Watson identified the same 703 cases, but also noted potential therapeutic options in 323 additional patients. Of those additions, 96 had not been previously observed to have an actionable mutation.
“To be clear, the additional 323 cases of Watson-identified actionable alterations consisted of only eight genes that had not been considered actionable by the molecular tumor board,” said William Kim, MD, corresponding author on the study and an associate professor of medicine and genetics in the UNC School of Medicine.
Kim went on to explain the main purpose of the study “was not designed to analyze whether or not this helps patients in regard to [the] outcome as defined by prolonged survival or treatment response.” While some of Watson’s findings were irrelevant (either because the patients didn’t have active cancer or had already died) the physicians of 47 patients with active cancer were notified of Watson’s findings.
The team’s research into cognitive computing is reminiscent of how AI’s machine learning capabilities have been used to understand suicidal behavior and to identify breast lesions that could develop into cancer. Although in terms of process, the research is most similar to astronomers using AI to find gravitational lenses — as both require culling through vast amounts of information that would overwhelm the human mind.
This isn’t the only way AI is being used in the medical field, and its applications will continue to expand as the technology is developed. Microsoft already has plans to use AI to find a cure for cancer, and AI has already proven capable of predicting heart attacks more accurately than doctors. These advancements hint at a future in which your trip to the hospital puts you face to face with a robot or supercomputer rather than a flesh and blood human.