DNA Dominoes

The idea of smart DNA tiles got its start five years ago at Caltech's Red Door cafe, when Winfree and Rothemund met to discuss Adleman's first DNA computing paper. The publication had set imaginations blazing throughout the world and across scientific disciplines. Were there other ways to compute with DNA? Could it beat silicon? Rothemund brought along a stack of papers showing "all the weirdest things that had been done with DNA." One of these was by Nadrian Seeman, a chemist at New York University who had created cubes, rings, octahedrons and other unlikely shapes from the DNA double helix. Winfree, who was working on a PhD related to artificial learning in robots, immediately saw a way that Seeman's strange versions of DNA could be used to compute.

Winfree's intellectual breakthrough was inspired by the theory of Wang tiles-a bit of recondite mathematics related to the patterns that can be created using squares with numbered sides. Like dominoes, the numbers on each Wang tile determine which other tiles it is allowed to touch. By carefully establishing these "matching rules," complex and interesting patterns can emerge as more tiles are added. But it's more than just a game of mathematical dominoes. Because Wang tiles carry both data (the numbers) and simple rules for combining it, mathematicians in the 1960s proved that the tiles could be used to add or multiply numbers. In fact, they showed that with the right set of these hypothetical constructs you could, in theory, do anything an electronic computer can-from playing chess to counting sheep. Winfree's big idea was a simple synthesis: use Seeman's DNA molecules as tiny real-life Wang tiles.

Applied to DNA computing, the strategy could sidestep one of the fundamental problems that has bedeviled the field from the beginning-too much lab work. While DNA computing is good at producing a vast number of answers quickly, things slow down when it comes to picking the right answers out of the mix. Take the traveling salesman problem originally solved by Adleman, in which the object is to find the most efficient route through seven cities connected by 14 one-way flights. Adleman created strands of DNA to represent each flight, then combined them in a test tube to generate every possible route.

Although the DNA in one-fiftieth of a teaspoon produced 100 trillion answers in less than one second, most of those answers were repeats-and most of them were incorrect. So Adleman's next task was to discard the wrong answers, something that could be done in a jiffy on a PC, but in Adleman's case required several dozen manual laboratory procedures. And that's where the trouble lies with most DNA computing schemes-each "operation" on the data means another time-consuming lab step.

The DNA tiles could solve that problem. Unlike the DNA used by Adleman in his original experiments that combined randomly, Winfree's tiles follow simple rules to get the correct result. "Ideally, you just put [the tiles] in the test tube and whammo!, you've got a right answer," says John Reif, a Duke University computer scientist.

Working with Winfree and Thom LaBean, a biochemist at Duke, Reif hopes to put the idea into practice by creating a simple molecular abacus out of DNA tiles. The goal is to add up binary numbers from zero to eight. With genetic letters standing in for 0s and 1s, the team has designed sets of tiles, each of which represents a possible column in an addition. Rules for combining columns correctly are coded into loose strands of DNA protruding from the sides of the tiles.

If all goes well, the experiment will generate several trillion multi-tile structures each of which has carried out an orderly addition of three binary bits. The scientists then will read off the results using standard methods for decoding DNA. The experiment underlines the potential power of DNA computers-massive parallelism and speed. Reif estimates that a single test tube of DNA tiles could perform about 10 trillion additions per second-about a million times faster than an electronic computer.

http://www.technologyreview.com/Biotech/12108/page2/

DNA Computing

Leonard Adleman sends his regrets. In an e-mail FAQ he uses to fend off journalists seeking interviews, the University of Southern California computer scientist and world-famous cryptographer who invented the field of DNA computing confesses that "http://www.technologyreview.com/Biotech/12108/are unlikely to become stand-alone competitors for electronic computers." He continues, somewhat apologetically: "We simply cannot, at this time, control molecules with the deftness that electrical engineers and physicists control electrons."

It was in 1994 that Adleman first used DNA, the molecule that our genes are made of, to solve a simple version of the "traveling salesman" problem. In this classic conundrum, the task is to find the most efficient path through several cities-given enough cities, the problem can challenge even a supercomputer. Adleman demonstrated that the billions of molecules in a drop of DNA contained raw computational power that might-just might-overwhelm silicon. But since then, scientists have run into tough practical and theoretical barriers. As Adleman and others in the field have come to realize, there may never be a computer made from DNA that directly rivals today'shttp://www.technologyreview.com/Biotech/12108/.

But that doesn't mean they've given up. Far from it. Although computer scientists haven't found a clear path from the test tube to the desktop, what they have found amazes and inspires them. Digital memory in the form of DNA and proteins. Exquisitely efficient editing machines that navigate through the cell, cutting and pasting molecular data into the stuff of life. What's more, nature packs all this molecular hi-fi equipment into a bacterium not much bigger than a single transistor. Viewed through the eyes of computer scientists, evolution has produced the smallest, most efficient computers in the world-and the beige-box set is hooked.

As Adleman now sees it, DNA computing is a field that's less about beating silicon than about surprising new combinations of biology and computer science that are pushing the limits in both fields-sometimes in unexpected directions. Scientists are still working hard on ways to tap the awesome number-crunching abilities of DNA for specialized types of applications, such as code breaking. But beyond that, the innate intelligence built into DNA molecules could help fabricate tiny, complex structures-in essence using computer logic not to crunch numbers but to build things.

Among the most promising of these new approaches are smart "DNA tiles" invented by Erik Winfree, a 30-year-old computer scientist at California Institute of Technology (see "100 Young Innovators," TR November/December 1999). Winfree's brainstorm is to create nanoscopic building blocks out of DNA that not only can store data but are designed-Winfree likes to say "programmed"-to carry out mathematical operations by fitting together in specific ways. Normally, DNA exists as two intertwined strands of the chemical letters A, G, C and T-the familiar double helix. But Winfree's DNA tiles are made by knotting together three or more of these strands, forming "tiles" about 15 nanometers (billionths of a meter) on their longest side. Taking advantage of DNA's ability to selectively recognize other strands of DNA, Winfree has "coded" the edges of these tiles so that they come together in just the right way to form tiny built-to-order structures.

In fact, programming DNA in this way could give chemists the kind of deft control "that may allow them to build more complex structures than any considered so far," says Paul Rothemund, a doctoral student in Adleman's USC lab.

http://www.technologyreview.com/Biotech/12108/

Slicing Up Silicon for Cheaper Solar

A California startup is cutting by half the amount of costly silicon used in solar panels.

Stripped-down solar: By cutting silicon solar cells into strips (left) and using molded plastic (right) to funnel light to them, Solaria is able to cut the cost of its solar panels. Credit: Solaria

Solaria, a startup based in Fremont, CA, intends to cut the cost of solar panels by decreasing the amount of expensive material required. It has recently started shipping its first panels to select customers. This spring the company will begin production of solar panels at a factory built to produce 25 megawatts of solar panels per year.

Current high costs for the type of silicon used in photovoltaics have significantly driven up the price of conventional solar panels. Solaria's cells generate about 90% of a conventional solar panel's power, while using half as much silicon, says Kevin Gibson, Solaria's CTO.

Ordinarily, the silicon in a solar panel spans its surface, collecting light from as much area as possible. But Solaria slices the silicon into thin strips and spaces them apart so that they only account for about half the panel's area. A clear molded plastic cover collects light from the entire panel and funnels it to the strips of silicon.

This approach saves money because the total costs of the molded plastic, other extra materials, and added manufacturing steps still are lower than the cost of the additional silicon used in conventional solar panels. Solaria also reduces costs by using manufacturing equipment already developed for the semiconductor industry, thus avoiding expensive customized equipment. Gibson says Solaria's first products will be economical enough to compete with panels produced by much larger companies, and that successive product generations will cost between 10 and 30 percent less than their competitors.

Silicon prices are high now. But the element is abundant, and already new facilities are coming on-line to produce more refined silicon. For Solaria is to be competitive in the long run, it will need to implement other cost-saving measures, especially improving the overall efficiency of its solar panels, says Tonio Buonassisi, a professor of mechanical engineering at MIT.

Such improvements are possible, Gibson says. For example, in conventional solar cells, wires for collecting current are placed on top of the cell, where they block some of the incoming sunlight. Solaria could place its wires between the strips of silicon, where they block no light. Because the wires wouldn't need to be made thin to avoid blocking light, they could be sized to collect electricity more efficiently

http://www.technologyreview.com/Energy/20412/