Ever vigilant to prevent the export of powerful computers, Customs inspectors once looked askance at Atari games. Someday soon, if a dream of computer-science visionaries pans out, the agents will have to be on the lookout for thimblefuls of genes. Researchers have long realized that nature computed eons before silicon chips did; she used molecules of DNA to store information, and biochemical reactions to carry out operations. As scientists bump up against the capacity and speed limits of electrons and silicon, DNA is looking better and better: recently the first DNA computer solved a classic mathematics puzzle. The idea "captures everyone's imagination," says computer scientist Richard Lipton of Princeton University, adding with professional detachment, "it really is totally cool."
DNA is suited for computing because "it is essentially digital," says computer scientist Leonard Adleman of the University of Southern California. While standard computers encode data as Os and Is, DNA uses four molecules, designated A,T, C and G. The idea is to string together the A's, T's, C's and G's in sequences that represent numbers or other information; the DNA strands react in a test tube, creating a molecule whose ATCG sequence is the answer. Though DNA computers perform an individual operation slowly (30 minutes, on average), the presence of trillions of the molecules means the biological computers do billions of operations at once-the ultimate parallel processor. An ounce of DNA could be 100,000 times faster than today's supercomputers.
Adleman used DNA to solve a version of the Traveling Salesman problem. A salesman has to visit seven cities, once each; each city is connected by a one-way road to only two other cities. Is there a shortest path the salesman can follow? Adleman represented each city by a strand of 20 random DNA letters. Each route also got a DNA name of 20 letters: the first 10 matched up with the city of origin and the second 10 matched the destination cit v. Adleman mixed billions of copies of each city DNA and each route DNA in a test tube; the strands hooked up in every possible way. With standard biochemistry, he fished out the strand that started with the DNA name of city of origin, ended with the final city and contained the other five DNA names in between. It took a week-instead of years for an electronic computer.
Lipton is using DNA computing to break cryptographic systems. He suspects that "a few months of biological computing" could crack even the National Security Agency's Data Encryption Standard DES), long thought unbreakable. What would a DNA computer look like? "Like a bunch of test tubes," says Adleman, with "material taken from one tube to another" automatically. Of course, DNA is not the most reliable molecule around. Its mutability is the basis for life on earth, and a stray cosmic ray could make it look dumber than a Pentium chip doing long division. Still, who would have thought, 50 years ago, that vacuum tubes could compute?