Designing Molecules

Drugs are molecular saboteurs. They exert their curative effects by gumming up the works of key proteins in the body. The compounds with the fewest side-effects are the ones that drop their monkey wrenches selectively, slotting into grooves on the surface of their target proteins — and no other proteins — as snugly as feet fit into socks.

But it's not easy to design drugs that choose their targets this efficiently. In fact, it's so difficult that drug companies have hardly ever tried. They've relied instead on trial and error, testing hundreds of potential drugs in animals to find a few that actually cure without killing. But these molecular crapshoots are terribly wasteful, which is why drug designers are today turning to a computer science known as bioinformatics to fuel their endless quest for newer drugs and better targets.

GeneFormatics of San Diego, Calif., for instance, uses bioinformatic algorithms to help drug companies predict the function of proteins encoded by newly discovered genes. It does this by comparing the new proteins to proteins of known structure, generating a "fuzzy" picture of what each looks like. That, in turn, suggests what their biochemical function may be — and how best to shut them down.

These fuzzy snapshots, however, aren't always enough. When they are active, protein molecules may double over or twist into radically different shapes. Understanding their dynamics can be crucial to drug design, and for this good computer simulations are invaluable. San Diego-based Structural Bioinformatics, for example, generates digital "movies" of how proteins writhe and twist when activated or ensnared by drugs, and identifies the small molecules that would best disable these moving targets.

The fact that computer-generated structures work barely a third of the time hasn't tempered the enthusiasm of drug designers. "Some knowledge is always better than no knowledge," says Arthur Olson of the Scripps Research Institute in La Jolla, Calif. And given the exponential growth of both computing power and data from the genome project, the technology can only get better.