Mayo And IBM Search For Personalized Medicine 2Mayo And IBM Search For Personalized Medicine 2

Hoping to customize medical treatments to individual patients--such as picking the best chemotherapy for a colon cancer patient with a specific genetic marker--Mayo Clinic and IBM are trying to apply pattern recognition and data mining to electronic records of about 4.4 million Mayo patients.

For the last three years, Mayo and IBM have been working to replicate into one database clinical data from five data sources, which include digitized patient files, lab results, X-rays, and electrocardiograms from Mayo's hospitals in Arizona, Florida, and Minnesota. In our lifetimes, each of us generates about 1 terabyte of medical data, from X-rays to test results.

Under an extension to that project disclosed Wednesday, Mayo and IBM will take that medical data and apply custom algorithms to identify patterns that could lead to better treatment decisions for individual patients and help researchers develop new therapies, says Drew Flaada, director of the IBM and Mayo collaboration and director of IBM life sciences.

Mayo and IBM are trying to apply to clinical information more-sophisticated versions of the pattern recognition and data mining that other industries use, such as by retailers to customize marketing or by banks to spot credit-card fraud.

For Mayo, the goal is to find patterns related to how patients--based on age, medical history, and other factors--respond to various treatments. This eventually could allow doctors to more closely match patients with treatment plans that work most successfully for patients with similar demographics and other characteristic. Flaada says such applications for practicing Mayo physicians might be available by the end of the decade.

For Mayo researchers, pattern-recognition tools also can be applied to one patient's data, such as looking for particular protein and genetic makeup, to help uncover current medical mysteries, Flaada says.

IBM products used in the project include: DB2 database software running on AIX-based P690 servers, which will be used for Mayo's centralized clinical dataware; WebSphere integration software; DiscoveryLink data-federation software; and DDQB data abstraction engine software. However, as the project progresses, Flaada expects software from other yet-to-be-identified technology vendors to also be used. Eventually, IBM hopes to make similar solutions based on the Mayo/IBM developments available to other medical researchers and providers, he says.

In parallel to the clinical data-mining and pattern-recognition project, Mayo Clinic will be the first medical institution to use IBM's Blue Gene supercomputer for molecular modeling for disease research. Flaada says data findings from the molecular research might in the future be part of the clinical data warehouse.