DRUG RESISTANCE: Virco Uses Computer Model to Identify Resistance Patterns

Scientists using a computer model of HIV have unraveled the mutation patterns in the virus' genetic code that make it resistant to the antiretroviral drug stavudine, the Wall Street Journal reports. The findings, presented yesterday by biotechnology firm Virco at the Fifth International Congress on Drug Therapy in HIV Infection in Glasgow, Scotland, are of "particular significance" because the resistant patterns for stavudine, manufactured by Bristol-Myers Squibb Co. under the trade name Zerit, have been "difficult to pin down" (Louis, Wall Street Journal, 10/26). A previous study, published by Stanford University, had identified nine mutations in the genetic code of stavudine, but was unable to predict resistance development to the drug. The new technique, an artificial intelligence program called "neural networks," identified a group of 26 mutations necessary for the virus to become resistant to stavudine -- a "more complex" profile than previously was suspected (Virco release, 10/26). Teun Grooters, director for antivirals clinical research for Bristol-Myers, said, "This model confirms that you need multiple mutations to get resistance to this drug." He added, "Once the model is fully mature it will give patients and doctors more confidence in determining the best treatment to have the best long term effect" (Wall Street Journal, 10/26). Traditionally, HIV drug resistance profiles have been determined using either genotyping or phenotyping techniques. Genotyping, although quicker and less expensive than phenotyping, is considered an "indirect assessment of HIV drug resistance," as the genetic code of the patient's HIV is read and mutations are interpreted, leaving room for mistakes. Phenotyping is a more direct measurement that examines the growth of the virus from a patient's blood sample as it is exposed to different drugs. Virco's "VirtualPhenotype," which uses the neural networks technology, combines both genotypic and phenotypic information to "generate large numbers of simultaneous equations: in essence trying different combinations of mutations to try to 'explain' the phenotypic changes in resistance." Each time the computer model runs, it learns and improves on its "predictive accuracy." Noting that the virtual model was a "very important development," Dr. Stefano Vella, president of the International AIDS Society, said that the technique is a "fascinating example of how we are using advances in a variety of different disciplines and applying them, in an integrated way, to accelerate progress in the HIV field." Dr. Brendan Larder, chief scientific officer at Virco and presenter of the study, noted that the induction of HIV drug resistance is "extremely" complex. "This approach harnesses enormous computing power to unravel the complex relationship between genetic changes and the resistant behavior of the virus," he added. Virco is also using the technology to identify resistance profiles and mutation patterns that cause viral resistance to new HIV drugs (Virco release, 10/26).