With the inexorable advance of high-throughput sequencing comes terabyte upon terabyte of data – certainly more than a human being can keep up with unassisted. Luckily, that same technological march forward means that we don’t have to cope with big data alone. Computer programs designed to identify specific genetic alterations or abnormalities can help – and when even those aren’t powerful enough for our purposes, highly sophisticated tools like cognitive computing hold promise for the future. A study from the University of North Carolina’s Lineberger Comprehensive Cancer Center (1) has investigated the ability of cognitive computing to identify treatment-relevant genetic mutations in cancer and discovered that it is a powerful ally.
“The idea to combine cognitive computing with genomics came out of a very practical need to keep up with the medical literature in general, as well as open, biomarker-selected clinical trials specifically,” says corresponding author William Kim. “As we held our weekly molecular tumor board, it became rapidly apparent to us that there had to be a more efficient and comprehensive way to go about the process than the ad hoc nature in which we were approaching every case.”
So Kim and his colleagues set IBM Watson for Genomics the task of searching the genomes of 1,018 cancer cases to find actionable genetic alterations. Their goal? To compare the computer’s performance with that of human cancer experts – in this case, UNC Lineberger’s molecular tumor board. The expert panel identified actionable mutations in 703 of the cases; Watson was able to spot them in 323 more, including 96 in whom no previous actionable mutations had been found. The computer also identified potential therapeutic options – clinical trials that applied to the patients’ specific mutations. “In general, we were quite surprised at how well Watson was able to inform us of the availability of open, -selected clinical trials,” says Kim. “For example, it informed us of a biomarker-selected clinical trial that opened only a week or two prior to our running the case through Watson.” Kim believes that cognitive computing will allow pathologists and laboratory medicine professionals to assess the relevance of a genomic alteration in a quick and comprehensive fashion. But there’s more: “Although not a reality today, cognitive computing may also hold promise in formulating preliminary reports that can be edited for placement into medical records.” He adds, “I have faith that cognitive computing in general has great promise for predictive biomarker discovery. However, this will require large datasets that are well-annotated with response as well as pre-treatment tumor genomics. We will get there, but it will likely take some time.”
Newsletters
Receive the latest pathology news, personalities, education, and career development – weekly to your inbox.
References
- NM Patel et al., “Enhancing next-generation sequencing-guided cancer care through cognitive computing”, Oncologist, [Epub ahead of print] (2017). PMID: 29158372.
