Mayo Clinic partners with IBM's Watson to improve clinical trial patient selection

At the Mayo Clinic’s Transform 2014 conference, Mike Rodin of IBM announced a new partnership with the clinic which will use the technology of Watson, IBM’s supercomputer, to match patients with the most appropriate clinical trials.

Rodin said a special version of Watson will be designed for the Mayo Clinic, which will include an integrated database of all clinical trial data from the Mayo Clinic, along with all public clinical trial information available, such as the data on www.ClinicalTrials.gov. Watson can then analyze patient data and compare it with trial criteria to determine which trial is best suited to each patient. With over 8,600 clinical trials conducted at the Mayo Clinic in 2013 alone, researchers face challenges in filling trials, and it is hoped the specially-developed system will improve the enrollment process.

Rodin referenced Watson’s famous debut on the television game show Jeopardy!, but explained that Watson does much more than simply answer questions. Instead, Rodin said Watson is designed to create a set of hypotheses and to define the probability of how well the technology has responded to a question; the system learns as it accesses more information. Rodin said in the case of clinical trial selection, or if the technology was applied in other health care settings, a physician would have the ability to look at the set of hypotheses Watson provides and make decisions based on the needs of the patient.

Watson’s technology could also aid in diagnosis or treatment selection, or deliver other forms of health care. Rodin said IBM has been working on developing technology to address these areas. He said the Watson technology could be developed to include vast information for making diagnoses by using all of the data available in PubMed and other clinical information and comparing that with patient information, including genomics and family history, then analyzing the data to generate a set of hypotheses about diagnosis or treatment. He said the top two or three hypotheses Watson returns may be the first ideas a physician had already considered, but the remaining hypotheses returned may provide insight into other possibilities a physician had not considered.

“Many times our interpretations are biased by our experiences,” Rodin said, adding that having information from outside of those experiences could prompt new thinking and better insight into diagnoses and diseases.

IBM is also developing Watson’s technology in other ways to benefit the medical community. Rodin said Watson was able to identify six new targets for the p53 protein for cancer treatment in just a few weeks, while researchers typically identify an average of one per year. To test the efficacy of the approach, Watson was instructed to search PubMed for data through 2003 only; the technology successfully predicted the next seven targets identified in following years.

Additional applications for the Watson technology are also underway, including a method of analyzing and condensing long and complex patient history data into a summary of information most important to a physician.

For more information:

Rodin M. Presented at: Mayo Clinic Transform 2014; Sept. 7-9, 2014; Washington, D.C./San Francisco.