While once the domain of fringe science fiction, the concept of using molecules to self-assemble and perform calculations like a computer chip has been the serious focus of the field of DNA computing for more than 10 years now. The work of DNA computing pioneer Leonard Adleman, a professor of computer science at the University of Southern California, led many in the field to believe that it might be possible for DNA to compete with silicon-based computing.
In the mid-1990s, Adleman's lab demonstrated the use of DNA to solve a version of the Hamiltonian Graph problem, which is similar to the traveling salesman problem. The results caused quite a stir. "When Adleman published his first paper in 1994, there was a bit of a gold rush and people jumped into the area, and very quickly a large number of papers appeared proposing various DNA-based solutions for hard [computational] problems," says Martyn Amos, senior lecturer in the department of computing and mathematics at Manchester Metropolitan University. "People thought if we can beat silicon, then we've found our killer app, and the field was driven by this search for the application of DNA-based computing that would establish its superiority over silicon."
But according to Amos, who earned his PhD with a dissertation on DNA computing from the University of Warwick in 1997, the issue of practicality in using molecules to do heavy computation is hampered by the inherent messiness of biology. "Small errors are very quickly amplified when you're using techniques such as PCR and affinity purification, which are fine for biologists for the level of accuracy they require, but not for a computer scientist," Amos says. "Tiny errors, which are imperceptible to the molecular biologist, are absolutely crucial to a computer scientist."
There are several methods for building a DNA computer, including logic gates using DNAzymes and oligonucleotides for input, DNA tiling, or PCR. DNA computers certainly use less power than silicon processors, are fast, and do not suffer from significant heat issues when compared to standard chips. However, scalability with a DNA computer is by its very nature a messy proposition. So while early DNA computing was characterized by a monolithic approach, like finding a statue inside the block of marble, many in the field have moved beyond the hope that someday biologically based microprocessors would surpass silicon chips to utilizing the computational potential of DNA strands to perform more basic logic functions.
One such researcher is Ehud Shapiro, a professor of computer science and biology at the Weizmann Institute, who is focused on practical applications for DNA computing that aim to execute relatively basic logical functions very quickly. Shapiro published a demonstration of the world's first programmable, autonomous DNA computing device — composed of enzymes and DNA molecules — back in 2001. To put it in perspective, a trillion of Shapiro's computing devices could fit in a single drop of water. This first device was able to perform simple tasks such as checking a list of 0s and 1s to determine if there was an even number of 1s. Then in 2004, Shapiro and his colleagues developed a DNA-based computer, complete with input and output modules that were capable of detecting cancer in a test tube and releasing a molecule to destroy it. The proof in the nano pudding was clear: these devices were capable of effectively interacting with a biological environment for medical and other purposes. "Our vision was that if we are to make a successful programmable, autonomous molecular computing device, they should be able to sense molecular symptoms of disease and analyze them on the input side and release a drug molecule on the output," Shapiro says. "So the vision for these kind of autonomous computing devices is that eventually they could be encapsulated in some appropriate packaging, delivered into a patient's body, and when traveling there they might be able to detect abnormalities that would require intervention."
Recently, Shapiro and his colleagues managed to develop a system that uses DNA strands capable of performing basic logic functions just like the countless number of components that make up any processing chip. They used DNA molecules to represent the various elements of a simple "if … then" train of deduction: All men are mortal. Socrates is a man. Therefore, Socrates is mortal. The device can successfully and reliably answer questions such as "Who is mortal?" and "Is Socrates mortal?" and other more complicated queries with additional rules and facts.
Shapiro's team also created a compiler to translate between a conventional programming language and a DNA computing "code." The compiler then translates the facts, rules, and queries into molecular representations, which in turn instruct a robotic system to assemble the logical deductions and produce a result. The results were encoded via flashes of green light emanating from a glowing fluorescent molecule that was revealed by an enzyme attracted to the site of the correct answer.
Drugs enabled with logic functions, or "smart drugs," have always been Shapiro's focus for DNA computing rather than attempting to rival a silicon processor. "It seems that [DNA computing] in general is gravitating towards the vision we proposed and slowly abandoning the vision of competing with silicon computers," he says. "I think [silicon computers] are doing pretty well and will continue to do so in the future, whereas there is an urgent need to have more intelligent drugs, and the potential of biomolecular computing is in this direction."
Fabrication, not computation
While Moore's Law has proven itself to be a reliable rule of thumb for the ever-increasing speed and complexity of processors, some researchers are looking to overturn the whole top-down assembly approach of making silicon wafers and transistors with increasingly smaller components. Currently, chipmakers have their sights set on methods to develop conventional lithographic and semiconductor fabrication techniques to make the components on chips smaller than 22 nanometers, the current target size of the industry. But some researchers have offered up proof-of-principle research that uses DNA molecules — not for computation, but as scaffolding or molecular-scale circuit boards upon which millions of carbon nanotubes or silicon nanowires can be deposited to form transistors for ultra-fast computation at the sub-22 nm level.
"Instead of conventional top-down assembly, where you keep having devices on the silicon wafer and you keep shrinking them smaller and smaller and smaller with each generation of technology, and as these features in the silicon wafer get smaller and smaller, it gets harder to do that," says Gregory Wallraff, a research scientist at IBM's Almaden Research Center in San Jose, Calif. "And it's this problem with the continual miniaturization that's making people look at other ways of doing it. And one way is to assemble components that are already at the molecular scale … so instead of shrinking things down, we'll start off with something small and assemble it the way we want."
IBM is working in conjunction with Caltech, where the techniques for creating so-called DNA origami were developed by Paul Rothemund in a process that makes individual DNA molecules self-assemble using a solution that mixes a long, single strand of viral DNA with a combination of various short synthetic oligonucleotide strands. The short segments can be used to fold the viral DNA into specific 2D shapes through base pair binding to create triangles, squares, or stars no wider than a DNA double helix. The team recently published research that used Rothemund's technique in conjunction with electron-beam lithography and oxidative etching to create binding sites on SiO2 and diamond-like carbon, both practical materials for chip making.
"The essential first step was to figure out a way to position [Rothemund's] origami structures on the silicon wafers in a precise way, to locate them precisely on the silicon wafer," says Wallraff. "The next step is to assemble the components which are things like silicon nanowires, quantum dots, carbon nanotubes — those would be the components of the transistors. The second step is going be as hard or harder." Beyond the challenge of getting these particles to assemble on the DNA templates, there is also the issue of defects caused by biology's inherent messiness.
So for now, the very flexibility and virtuosity that gives DNA the potential to aid in computing remain the biggest hurdles facing its practical application.