University of Rochester

New Formula Brings Dream of Ultrapowerful DNA Computers Closer

August 5, 1996

Ever since the very first computer five decades ago, computers have been shrinking, from room size to laptop size to the fingernail-size chips that rule our lives. Now, computer scientists are intrigued by the promise of a whole new kind of computer, made of DNA. Just a test tube full of these strands of life could process more information than all of today's supercomputers combined, yet would use only a billionth of the energy a laptop needs.

There's just one truly gigantic problem: Problems run by cabinet-size supercomputers today would demand a mass of DNA the size of the sun itself.

Now a University of Rochester researcher has proposed a new method that slashes the volume of DNA needed for most problems to a single test tube, an important step in the effort to make DNA computers a reality. In computer simulations, the new formula can tackle a problem many billions of times more complex than those solved by other proposed DNA computers, thanks to its sleeker memory requirements. The algorithm was created by computer scientist Mitsu Ogihara, one of only a handful of researchers nationwide working to develop a DNA computer.

For years computer scientists have been fascinated by the potential of DNA, the genetic blueprint present in all organisms, as a compact and efficient source of memory in biological computers. Indeed, seven feet of the material coiled up inside each of our cells serves as the instructions for nature's most complex program -- the human mind and body. Why not take that ability to encode information and put it to work on other problems?

First, computer scientists must find the "right" DNA for any given problem. Currently there is no good method to winnow out "good" computing DNA -- the strands with the correct sequences of the four chemical bases (thymine, guanine, adenine, and cytosine) that would serve as the computer's memory. So researchers generate all possible strands of a given length of DNA, producing mostly "waste" DNA, and then extract the few strands they'll ultimately need. For complex problems such as those commonly solved on today's computers, that would mean, literally, tons of extra DNA, an amount impossible to produce and manage. It's this problem that Ogihara's work addresses.

"In my algorithm, instead of creating all the candidates at the beginning and then finding the ones we want, we build the strands gradually, continually eliminating strands that we cannot use," says Ogihara. "In this way we cut down dramatically the amount of DNA needed.

"Computer scientists always juggle space and time: the bigger the memory, the faster the computer. Usually we try to increase the size of the memory to make the computer faster, but here we must decrease the memory. Even with a smaller memory, a DNA computer would still work much more quickly than a conventional computer."

A DNA computer works in much the same way as a conventional computer, where information is stored as "zeroes" or "ones" based on the spin of magnetic particles on disks. In a DNA computer, the sequence of bases in a molecule of DNA encodes information that could be accessed and altered with cut-and-paste biochemical techniques widely used today.

The potential benefits of a DNA computer are astounding. One pound of DNA could have the capacity to store more information than all the computers ever built, and a test tube has the capacity to store ten million times more information than the most powerful supercomputer. Ogihara believes DNA will prove useful on problems that now require vast networks of computers, such as forecasting the weather, designing airplanes, or cracking complex security codes.

Richard Feynman proposed the idea of a biological computer in 1959, and in 1994 Leonard Adleman of the University of Southern California first demonstrated the concept by using DNA to find the quickest route between several cities. Since then nearly a dozen algorithms have been proposed by groups at institutions like Rochester, Princeton, Penn State, Duke, and the University of Wisconsin. Much of the work, like Ogihara's, is theoretical. Besides the space problem, there are other daunting issues: DNA usually decays after several days, and its replication is not always completely accurate. Nevertheless, Ogihara is upbeat.

"I think DNA computation is quite promising," says Ogihara. "By the end of this century, I hope to have a simple DNA computer capable of adding two-digit numbers. That may seem like a minor development, but it's an important step toward DNA computers that can carry out complex calculations." sb