Feb 23, 2005 (12:02 PM EST)
Microsoft Brings Software Know-How To HIV Research

Read the Original Article at InformationWeek

Microsoft on Wednesday outlined its latest efforts in applying data-mining algorithms, machine-learning techniques, and computer modeling to the search for more effective HIV vaccines. Some of the technology being used will be included with Microsoft's forthcoming SQL Server 2005 database.

Microsoft researchers are working with doctors and scientists from the University of Washington and Australia's Royal Perth Hospital in an attempt to identify patterns of genetic mutation in the human immunodeficiency virus and in the immune systems of HIV patients. Lab tests began this month on vaccine models developed with Microsoft's help. The company demonstrated how its technology is being used at the Conference on Retroviruses and Opportunistic Infections in Boston.

Microsoft has been working in the area of HIV research for about 18 months, says David Heckerman, lead researcher and manager of Microsoft Research's machine learning and applied statistics group. After an initial project to develop a preliminary vaccine "design" using clustered Windows systems, the work has branched into three research initiatives, one of which uses graphical-modeling capabilities to ship with SQL Server 2005, formerly called Yukon and in beta testing now. That part of the project involves creating graphical models to study how HIV mutates within an infected person and how mutations are influenced by a person's immune system.

"We're very proud to see [our software algorithms] appear in this Microsoft product," Heckerman says. Microsoft's 5-year-old SQL Server 2000 database shipped with data-mining algorithms, too, but SQL Server 2005 will come with "many more" algorithms and more sophisticated algorithms, Heckerman says.

The two other initiatives involve software developed by Microsoft Research that has yet to appear in Microsoft products. One applies machine learning to understand how the immune system reacts to foreign proteins, called epitopes, that are found on infected cells. The software employed is similar to Microsoft's spam-filtering technology, but where algorithms attempt to sort out good cells from bad cells. "You're trying to find needles in a haystack," Heckerman says.

The third area of research involves applying an algorithm called Epitome that looks for ways to compress epitopes so that more effective vaccines can be developed. Epitome was originally developed by Microsoft researcher Nebojsa Jojic to compress images.