IBM's Watson cognitive computing technology has helped doctors, loan officers, corporations, and even military veterans find answers to complex questions. The next big challenge for Watson is helping researchers explore the unknown.
IBM announced Thursday that its Watson Discovery Advisor technology is now available as a cloud service. Backing up Watson's value in research roles, Baylor College of Medicine and IBM published this week a peer-reviewed study that came up with six promising paths for cancer research with the aid of Watson Discovery Advisor.
As part of Baylor's research, Watson analyzed more than 70,000 scientific articles related to p53, a protein that has been linked to many cancers. Automated analysis carried out by Watson helped Baylor biologists and data scientists identify six proteins that modify p53 and that should be targeted for new research. Most important, the discovery was made in a matter of weeks, according to IBM.
"In the life sciences industry at large, researchers typically come across one of these target proteins per year," said IBM Watson VP John Gordon. "Baylor working with Watson found six targets, and the first two that they've taken into wet labs have been validated, so they're outpacing the industry."
The pace of biomedical research has greatly accelerated in recent years with breakthroughs in speedy, low-cost DNA analysis. What's more, leading pharmaceutical companies such as Pfizer, Novartis, and many others are routinely finding correlations among genomic, clinical trial, and de-identified electronic medical records. Where Watson stands out, according to IBM, is its cognitive computing understanding of language and chemistry to make human-like leaps in understanding at computer-analysis speeds.
"People are already finding correlations among disparate sets of data, but because Watson can understand concepts and interpret the direction of research, it can uncover relationships that are more subtle," Gordon told InformationWeek in a phone interview.
For example, research papers don't just declare whether proteins are related to p53 or not, drawing simple, binary conclusions; they explore whether these proteins accelerate or inhibit mutation and what chemical processes they might catalyze. Now multiply the challenge of absorbing these subtleties by 70,000 research papers.
Click here to continue reading on EE Times' sister site Information Week.