Computers shape AIDS-drug search.

Computers shape AIDS-drug search

Using the rules of biochemistry, researchers over the years have developed hundreds of compounds to fight bacterial diseases. But attempts to expand the tiny arsenal of antiviral drugs pose greater challenges, in part because viruses -- including the AIDS-causing HIV -- replicate within the cells they infect. Rather than focusing on complex viral biochemistry, some scientists have turned to viral geometry -- and tailor-made computer programs -- to identify new antiviral weapons.

Last week, researchers described the first fruits of that approach as applied to HIV. Led by chemist Irwin D. Kuntz Jr. of the University of California, San Francisco, the team used specialized software to search a computer database depicting structural images of thousands of existing drugs, looking for molecules with just the right shape to bind and inhibit the activity of a key HIV enzyme.

Haloperidol -- a long-established antipsychotic drug -- turned up unexpectedly as the best fit.

The researchers then moved from the computer to the lab, showing that haloperidol indeed binds to purified HIV protease and, at high doses, slows HIV infection in cultured human lymphocytes. Kuntz reported the results at a research conference on AIDS held at the National Institutes of Health in Bethesda, Md.

This discovery offers no clinical benefit in itself, he emphasizes, because inhibiting the HIV enzyme would require 1,000 times the standard haloperidol dose -- enough to kill any prospective patient. But the work does point to a new and relatively rapid method for identifying potential AIDS drugs, Kuntz says. If drug companies were to use specialized software to identify drug molecules with shapes that could lock onto the valleys, pockets and other surface peculiarities of HIV constituents, "all [they would] have to do is test the drugs to know quickly if any show promise of treating AIDS," he says.

To search for the perfect match between a drug and a targeted viral component, Kuntz's software system first constructs an inverse image of the target's shape. The surface of a drug molecule must match this inverse image in order to bind and inhibit the activity of the designated component.

The team targeted the viral enzyme known as HIV protease because other scientists had recently succeeded in crystallizing it, allowing detailed analysis of its structure. In addition, previous studies had shown that HIV, when replicating inside cells, relies on protease as a molecular scissors. The enzyme first snips itself from cellular material, then cuts out other proteins essential for development of the mature virus. Compounds that bind to HIV protease block the enzyme's activity, halting maturation of the next generation of viruses and leaving them vulnerable to immune attack.

A scan of three-dimensional computer images for 10,000 drugs revealed haloperidol's surprising potential. And unlike the easily digestible and short-lived peptide compounds already known to inhibit HIV protease, haloperidol's chemical structure may allow it to remain active far longer in the body, Kuntz notes.

"We got very excited," he says. "We realized we didn't have to change the structure from what we saw on the [computer] screen."

Kuntz told SCIENCE NEWS his group has developed a haloperidol derivative that blocks HIV protease at lower -- but still lethally toxic -- concentrations. The researchers are still struggling to modify the drug for safe use in AIDS patients. They have also begun work to identify and modify other potential weapons against HIV.

In a separate effort, Krzysztof Appelt of Agouron Pharmaceuticals Inc. in La Jolla, Calif., is using similar software to further expand the arsenal of nonpeptide compounds that inhibit HIV protease.