Computer science is giving scientists new ways to look at the virus that causes AIDS (acquired immune deficiency syndrome), perspectives that may help efforts to develop an effective vaccine and other medicines, according to the head of Microsoft's research arm.

"It's really focused on new ways of thinking about how to describe and analyze systemic activities within a cell," said Rick Rashid, Microsoft's senior vice president of research, during an interview in Singapore.

Human immunodeficiency virus (HIV) causes AIDS, a disease that weakens a person's immune system leaving them vulnerable to infections and other diseases that would rarely threaten a healthy person. The U.S. Centers for Disease Control estimates that 1.1 million Americans are infected with HIV, with 20 percent of those people unaware they are carrying the virus.

Computers and software are already in widespread use by AIDS researchers, but computer science is proving useful in other ways, as well.

"We've been finding that a lot of underlying computer science theory, especially computer science languages, can actually be used to describe cell processes, and then the mathematics that we use to analyze programs can also be applied to analyze cell activities because there's an underlying mathematical relationship," Rashid said.

"It's opening up peoples minds to how computers can help them, not just to do their work better, but how the underlying theory and underlying computer science changes the way they look at their problems," he said.

Since 2005, Microsoft has sought to apply machine-learning techniques, including technology used in spam and antivirus filters, to AIDS research. The goal is to find genetic patterns in HIV that can be used to "train" the human immune system to fight the virus. In particular, Microsoft has looked for ways to track how HIV mutates to evade the human immune system.

To that end, the company last year released software code for four tools developed by Microsoft Research to help researchers analyze the genetic makeup of HIV.

"The idea is that because the genome is basically digital, it can be described as a string and analyzed as a string. It opens up an opportunity to think about a lot of problems in that space as data mining or machine-learning problems," Rashid said.

The IDG News Service is a Network World affiliate.