I.B.M. Using Bits of Watson for Drug Research

When I.B.M.’s Watson computer beat two human “Jeopardy!” champions earlier this year, it was a triumphant demonstration of the company’s technology. It was great for Big Blue’s image, but it was not a moneymaker on its own.

Yet that process is under way. Watson was a bundle of advanced technologies, including speech recognition, machine learning, natural-language processing, data mining and ultrafast in-memory computer hardware. They have been under development at I.B.M. for years, and were pulled into Watson.

The ingredients that went into the Watson arsenal are steadily finding their way into I.B.M. products. For example, WellPoint, the big health insurer, is trying out a system that uses Watson-style software to reduce redundant medical tests.

The latest entry is being announced on Thursday, I.B.M.’s Strategic Intellectual Property Insight Platform. Clearly, the Watson branding team did not work on this name.

But then again, this is not for television, where Watson performed, it is for major corporate customers seeking competitive advantage. The technology, sold as a cloud-based service, is the result of several years of joint development between IBM Research and four companies — AstraZeneca, Bristol-Myers Squibb, DuPont and Pfizer.

The insight platform uses data mining, natural-language processing and analytics to pore through millions of patent filings and biomedical journals to look for chemical compounds used in drug discovery. It searches for the names of compounds, related words, drawings of the compounds, the names of companies working with specific chemicals and molecules, and the names of scientists who created the patented inventions. It does its work quickly, retrieving information on patents in as little as 24 hours after a filing.

“It provides a landscape that shows who is working with what chemicals and drugs,” said Chris Moore, head of business analytics and optimization in I.B.M.’s global services unit.

The technology, Mr. Moore said, can be applied to everything from product strategy to recruiting to patent enforcement.

As a byproduct of its research, I.B.M. is also adding to a vast, searchable chemical database housed by the National Institutes of Health.

The company is contributing more than 2.4 million chemical compounds extracted from 4.7 million patents and 11 million biomedical journal extracts from 1976 to 2000.

The information was all published, but often in costly scientific journals or buried in the mountains of patent filings. It was so difficult to access that it was, for all practical purposes, inaccessible.

“It’s a nice contribution to the field of open chemistry — and that’s a growing trend, inspired by and similar to open source software,” said Marc C. Nicklaus, head of the computer-aided drug discovery group at the National Cancer Institute, which is part of the National Institutes of Health.

I.B.M.’s data contribution, it seems, is both generous and calibrated. The chemical compound data from patents goes from 1976 to 2000. So most of the data will be on patents that have already expired, useful for scientific research but far less useful commercially. The latter, no doubt, will be of greatest interest to I.B.M.’s paying clients.