company had claimed. There remained major questions: Was it small enough to get inside the brain, something Searle’s compound had not been able to do? What of possible allergic reactions? VX-478 had chemical groups similar to those thought to cause some people to react violently to the antibiotic Bactrim. Carl Dieffenbach, the NIH point man for assessing new AIDS drugs, mused to a reporter, “I’m not sure their patent is as secure as they think it is.”
Boger weighed these questions only to brush them aside, especially the last. As part of its due diligence, Wellcome had satisfied itself that Vertex’s patent claim was clean. Its lawyers had won—and held on to—the rights to AZT through ten years of intense legal strife, providing an intimidating ally. A few months later, during a November conference call with Wall Street analysts, Searle announced it was stopping development of its protease inhibitor. Two clinical trials had showed no indication of antiviral activity, and Searle researchers believed that the compound was soaked up by a blood protein and removed by the liver. Boger was relieved to learn that the problem was unique to Searle’s molecule, and he considered himself fortunate to have the threat of a patent war suddenly diminished.
Murcko, as usual, thought well beyond the problem at hand. With the genetic material from Charlie Rice, the main scientific challenge was to discover how the hepatitis C protease worked and how to disable it. But Murcko had a secondary question, one more central to Boger’s mission: How do you evaluate new targets without knowing their structures? Could you predict, say, degree of difficulty? When was it wise to invest in new projects and when wasn’t it? “Up to now at Vertex, we said we either want to have a crystal structure available or we want to know that we can get there first,” Murcko says. “But sometimes maybe you wouldn’t be able to get the crystal structure as quickly as you’d like. Maybe if you hada model it could help steer you away from certain projects. Or say this one isn’t going to appear so easy.”
Murcko did the experiment. Having no new positions but recognizing the uncommon gifts of a recent Harvard postdoc named Paul Caron, he hired Caron as a temp. Caron was advanced in thinking about gene sequencing and physical similarities among proteins, which fold into spirals and loops and cascading sheets according to the ordering of the amino acid residues out of which they’re made, following the text encoded in their DNA. In other words, if you had the genetic code for a target you might be able to model its active site prospectively, by mapping it against other known proteins. Sitting at a workstation in the modeling room, with a second PC at his side so he could calculate atomic charges and distances, Caron lined up the structures of the few other viral and mammalian proteases that were available and quickly noted that HCV lacked the usual cleft that caused the binding site to be buried in a pocket. “What we saw was virtually a bowling ball,” he said. “Very smooth. All the big loops that come around and make the channel weren’t there.” The target was going to be far more difficult than anything they’d done before.
One of the company’s advisors, Harvard structural biologist Steven Harrison, categorically dismissed the model, saying it couldn’t be true. Others confessed equal doubts. No one wanted to stop the project, but Murcko and Caron, knowing that the featureless active site presented a steep challenge to design, wanted people to be realistic. “The question at the time, once we had this model, was, ‘Do we continue?’ ” Caron recalls: “It’s going to take a relatively large inhibitor to get enough binding energy, and it’s not going to be an easy thing. We said, ‘This isn’t going to be HIV again. This is going to be a long, hard project.’ ”
Thomson and his protein group also were learning that HCV