"If you start with compounds that are FDA-approved, it may be a faster way to find good drug leads," says Rommie Amaro, who specializes in pharmaceutical and computer sciences at the University of California at Irvine. "There's a long process to get a drug reviewed, and the molecules have to be metabolically okay for people to ingest. So instead of starting with random leads from a chemical library, if you start with compounds that are FDA-approved, you could already have the more harmful compounds weeded out."

The process gave Dadon 15 hits, all with a higher binding affinity for H1N1 than any of the antivirals already approved for use against flu. Because all 15 of those compounds had a single substructure in common, Dadon looked for molecules already being produced by chemical and drug companies that contained that substructure. He found six of them, and all are currently being tested against H1N1 by collaborators in Australia.

The flu virus mutations known to resist antiviral drugs appear to occur in an altogether different binding site than the one Dadon and his colleagues discovered. Such a distinction is important, especially as reported cases of H1N1 resistance to Tamiflu are on the rise. If any of the researchers' six molecules prove successful, the resulting drug could provide a second line of attack to be used where current antivirals fail.

McCammon used an earlier version of the same method to discover novel inhibitors for a key HIV enzyme. But the technique is still relatively obscure in the drug-discovery world. "The relaxed complex method is not widely used because of the amount of computing time that it requires," says Wilfred Li, a bioinformatics specialist at UCSD's San Diego Supercomputing Center.

Li also notes that such studies are quite computationally expensive. "This study really stresses the need for a better computing infrastructure so more proteins can be studied in this fashion. These techniques can be used to develop new, interesting, and more potent inhibitors."