By Maxine Bookbinder

February 5, 2016 | IBM Watson Health and Boston Children’s Hospital have created a joint venture to apply Watson’s cognitive platform to help BCH clinicians diagnose and treat rare pediatric diseases.

In the five years since it made its debut on Jeopardy!, Watson has emerged as a valuable healthcare partner. In April 2015, IBM launched Watson Health and the Watson Health Cloud platform, allowing doctors, researchers, and insurers to obtain and organize new information from the growing accumulation of clinical, research, and social health data shared daily. “This is the intersection of science and technology,” says Steve Harvey, vice president of IBM Watson Health.

Watson learns about healthcare by “reading” the medical literature, searching for key terms like biomarkers, symptoms, diseases, and treatments, and learning how they relate to one another. Approximately 150,000 medical articles were published just in 2014, “a voluminous amount for any human to read,” says Harvey. But Watson can read tens of millions of pages in minutes, depending upon content and data, and can sometimes extract surprising insights like new biomarkers.

“Watson is very good at extracting specifics,” says Harvey. “We use technology to identify the most likely culprits. Technology is very good at taking a task and making it replicable. We don’t worry about human error.”

Watson’s cognitive base is always increasing, building upon past collaborations. So far, IBM Watson Health has partnered with Memorial Sloan Kettering Cancer Center, the Mayo Clinic, New York Genome Center, the Cleveland Clinic, and the University of Texas MD Anderson Cancer Center. It has also teamed with 16 cancer institutes in Japan, Canada, and the U.S., analyzing genomic data and helping oncologists create personalized treatments. All literature has been in English, but Watson is being taught to read other languages, including Spanish, Arabic, and Japanese.

BCH became involved with Watson Health in September 2015, when it announced a partnership with IBM to create OPENPediatrics, an online platform that will help pediatricians obtain the information they need to treat patients worldwide. Now, the collaboration is expanding to learn whether Watson can help diagnose rare diseases within BCH’s own patient community.

In an initial collaboration centered on nephrology, Watson will be trained by reading volumes of relevant medical literature on steroid-resistant nephrotic syndrome (SRNS), a rare, genetic form of kidney disease diagnosed in young children that often does not respond to immunosuppressive treatment, resulting in kidney failure and requiring transplantation―only to return in the new kidney. By scanning the literature, Watson will learn specific medical vocabulary, syntax, and language needed to make sense of individual cases of the disease. Friedhelm Hildebrandt, M.D, chief of the Division of Nephrology at Boston Children’s and Claritas Genomics will collaborate on the kidney project.

Once Watson has been educated in the subject, the plan is for experts at BCH to feed the platform genomic sequencing data from former patients, creating a Watson-based cognitive system that will help clinicians interpret a child’s genome sequencing data. Researchers typically do whole exome sequencing on both patients and their parents, to help identify possible mutations causing the disease.

“Reading sequence used to be a problem,” says Harvey. “Now we can read an entire genome of cancer in 37 hours. The question is what to do with the information.”

Drawing on this patient-specific genomic data, Watson will search the medical literature to identify anomalies and mutations that may be responsible for the disease. “We also want to know what percentage of the population has it and the age of onset,” says Harvey. The collaboration is still in proof of concept stage, scheduled to be completed in February.

Harvey says that, based on success with this project, “we are indeed planning to scale the solution to other rare diseases in the future. We will know we're successful when the doctors at Boston Children's tell us that Watson is an essential tool assisting their analysis efforts.” However, effective treatments, as with SRNS, can be elusive. In the U.S., a condition is considered “rare” if it affects less than 200,000 people.  One in 10 Americans is diagnosed with a rare disease; 80% are genetic. According to the Global Genes Project, half of those diagnosed are children, 30% of whom will die before age five and leaving parents to fear siblings will also become ill.

Approximately 7,000 rare diseases have been identified; new ones are discovered almost daily but remain undiagnosed due to a deficiency of clinical research. Funding, often directed toward well-known and widespread illnesses, remains limited. As the cost of genetic sequencing continues to drop, more data will be collected and published, further overwhelming researchers and clinicians already faced with a capacious amount of untapped information.