DNA - the chemical star of the O.J. Simpson trial - is showing remarkable mathematical capability. Its capacity to store and manipulate information lets it function as a computer. In other words, it lets computer scientists do math in a test tube that would be difficult to do on an ordinary computer.
This challenges our notion of what constitutes a computer. It also hints at the possibility that the computer-like qualities of DNA may play a role in the evolution of organic life.
Leonard Adleman of the University of Southern California in Los Angeles kicked off this research last November by showing how to solve a simple example of an inherently difficult graphical problem. Since then, other scientists have jumped on the idea. Richard Lipton of Princeton University in New Jersey, for example, recently showed theoretically how a molecular computer might solve a complex type of algebraic problem.
This research is so new it's impossible to tell where it will lead or what practical use may emerge. As David Gifford of the Massachusetts Institute of Technology points out, "the potential is really great." But, so far, it's only solved "one toy problem of one type."
That type of problem involves finding a path through a graph with a given number of points and edges connecting them. It could be a map showing cities with connecting airline routes. The trick is to find a route that begins at a designated city, ends at a different designated city, and visits each city just once. Dr. Adleman used only seven cities to keep the problem simple.
DNA encodes information as sequences of four types of chemical units called nucleotides and designated A, C, G, T. Adleman represented the cities as random sequences of these four letters - sequences that were 20 letters long. These were DNA "names" for the cities, so to speak. He represented the connecting routes by other 20-letter strands of DNA. Each of these was made up by taking the 10 letters at the end of the DNA name of the departure city and attaching the 10 letters at the beginning of the DNA name of the destination city.
With genetic engineering techniques, scientists can fabricate any DNA sequences they want. They can combine them, restructure them, and identify and extract any sequence. In his experiment, Adleman mixed his various DNA strands together. They combined to form longer DNA strands made up of several DNA names for various cities. He then was able to extract the strand that began with the designated departure city, ended with the designated destination city, and contained the names of all the cities just once. This represented the solution to the problem.
This shows up the strength and the weakness of molecular computing. An electronic supercomputer that does trillions of operations a second can solve many problems in a flash. It can take hours or days of chemical manipulations to solve problems with DNA computing. But the cities problem gets difficult fast as more cities are added. It soon overwhelms an electronic supercomputer that, at best, explores a few possible solutions to a problem at a time. With trillions of DNA molecules, each acting as a processor, a molecular computer generates billions of possible solutions simultaneously. Thus, when problems can be broken down into many simple steps that can be taken simultaneously, a molecular computer may have an edge.
It remains to be seen if this capacity will have practical applications. Meanwhile, exploring it will undoubtedly yield new insights both into computing and into life processes themselves.
Adleman's technique is to generate all possible solutions and then pick out the ones you like. In organic evolution, the same molecule - DNA - also generates many possible capabilities and nature picks out what it "likes." Biologists now may ask what viewing evolution from a computing perspective can tell them about the rise of Earthly life.
Adleman may have worked with a "toy" problem. But he has raised awesome questions.