This article is reproduced with the permission of New Scientist for exclusive use by Nova users.

DNA processors cash in on silicon's weaknesses
04 August 2006
NewScientist.com news service
Tom Simonite

Enlarge Sequence Spotter

It seemed as if we were on the verge of a huge breakthrough in computing.

Twelve years ago, Leonard Adleman used a mixture of DNA molecules in a test tube to solve a simple mathematics problem. Adleman, a computer scientist at the University of Southern California, had realised that if a problem could be encoded on DNA strands, molecular tools used by biologists to manipulate DNA could instead be used to crunch numbers.

The idea caught on fast, and there were high hopes that DNA computers could ultimately compete with and even surpass the processing speeds achieved by electronic computers. Researchers created various test-tube DNA mixtures that could solve other mathematical problems, and the approach was even used to crack IBM encryption codes.

But then, just as quickly, the dream of using DNA to rival electronic computers faded. The problem, says Adleman, is that DNA molecules are simply not as well-behaved as the electrons in conventional computers. "Only 90 per cent of our molecules would do what they were supposed to," he says, which produced unacceptably high error rates. "That was the limiting factor when it came to building bigger, faster systems." Silicon appeared untouchable after all, and since 2002 Adleman and many others have drifted away from the field to focus instead on using DNA to build precision nanostructures.

However, not everyone has given up on solving problems with DNA computing. A small number of devotees around the world have simply decided to switch the kind of problems they tackle with this technology. Leave the number-crunching to silicon, there is one area where DNA wins hands down: if you want to investigate biological systems, you need biological hardware.

"Biologists deal with DNA in the lab all the time, and DNA computation is perfect for making that job a lot easier," says virologist Joanne Macdonald of Columbia University in New York. She believes the technology could dramatically speed up the identification of virus strains and the genetic markers used to locate genes linked to disease.

Biologists are increasingly using "microarray" DNA chips to identify pathogens. Various known DNA fragments are attached to the chips, and samples of the DNA to be identified are tagged with a fluorescent molecule and washed over the array. Only those sequences that bind to the complementary "capture" DNA sequences are left behind. An automatic reader then detects their fluorescence and works out which sequence they bound to.

However, the technology is relatively slow, says Macdonald. "Everyone thinks microarrays are great, but you have to use formaldehyde and heat and wash them before you get a result. That's a 2 to 3-hour process." A DNA computer could do the same job much faster, she says.

Macdonald has been working with Milan Stojanovic, also at Columbia, who in 2003 built MAYA, a DNA computer that could play a stripped-down version of noughts and crosses (tic-tac-toe) against humans and never lose, (New Scientist, 23 August 2003, p 19).

Macdonald is using MAYA's logic gates, which comprise small wells containing DNA-based complexes, to create a computational tool for spotting particular sequences. "Our logic gates produce results in 15 minutes," she says.

All DNA computing relies on the ability of complementary sequences to pair up - the chemical "letter" A pairing with T on the opposing strand, and G with C. The test strand of DNA forms the input in the logic gate, whose DNA apparatus binds only to a particular DNA sequence (see Diagram). Only if the gate detects the target sequence does it fluoresce.

Macdonald has already used these logic gates to distinguish between different strains of West Nile virus, using a set of six gates, each designed to identify a different variation of one West Nile gene. She has also pooled four different gates in one to distinguish sequences that differed by just one DNA letter. These one-letter differences, known as single nucleotide polymorphisms (SNPs), are used to detect particular genes that code for disease, or even identify human remains.

"But the real strength of the gates is the next step," says Macdonald. By combining different logic gates it should be possible to quickly answer questions that would usually require computer analysis, such as whether a sample of bird flu virus contains a combination of sequences that make it particularly virulent, for example - different gates could be used to spot different sequences in parallel. Since each droplet of fluid contains over 1 trillion detector molecules, the devices can also be made extremely small. "It would be great for fieldwork where it's hard to lug all the normal array-reading equipment around," she says.

Chemists at the University of California, Berkeley, are also investigating DNA computers for performing simple biological calculations. They have developed a chip that could ultimately be used to "process" human genes to look for significant combinations of SNPs.

"DNA computing has so far been very much at the test-tube level," says Will Grover, a chemist who has recently left the group to pursue his own research at the Massachusetts Institute of Technology. "We took a lesson from history - electronic computing didn't really take off until the development of the integrated circuit." His device consists of tiny channels and valves that can form circuits like those on an electronic chip, with little need for outside control.

Every channel contains a bead coated with a known sequence of DNA. The sample DNA is circulated through each channel in turn, controlled by the valves. In each one, matching sequences bind to the bead, and the rest are washed away. The captured sequences are heated to break the weak bonds between strands, releasing them from the beads. At the end of the process, only if the original sample contains all of the sequences on the beads will any DNA remain. This might allow you to look for samples that contain certain sequences but not others, to spot, say, a particularly virulent combination of genes.

"Recirculating the DNA computing components makes computation much faster," says Grover. The team is interested in using the technique to look for particular "haplotypes" - collections of SNPs on a given chromosome - associated with genes that code for diseases. They hope soon to test the approach on DNA from a real organism.

Ultimately, such biological computers could move out of the lab and into a very different environment. Ehud Shapiro at the Weizmann Institute of Science in Rehovot, Israel, is aiming to put computers into patient's cells. In late 2004 he revealed a system called the "DNA doctor", a mixture of DNA and enzymes that, in a test tube, can detect a diagnostic indicator of prostate cancer and release an active drug compound in response.

Shapiro's original aim was to create a DNA "Turing machine" - a processor capable of verifying mathematical propositions - but like most in the field he has since abandoned any ideas of DNA number-crunching. "We found a more useful application for a simpler system," he says. His team is now attempting to get the DNA doctor working inside a living mammalian cell.

Shapiro says artificial DNA computers could lead to new medical treatments. "A DNA computer could potentially circulate in the blood and reach every cell in the body. There is no way to do that without computation systems made from biological components."

From issue 2563 of New Scientist magazine, 04 August 2006, page 24-25

For the latest from New Scientiist visit www.newscientist.com