DNA computers

Besides the self-assembling structures we discussed recently, DNA’s base-pairing lets it function in a weird kind of computation. Most versions of such computation involve DNA strands containing short sequences that can base-pair to complementary sequences on other DNA strands. “Computation” occurs when the DNA strands are put together in solution, and rapidly “find” all their best-fit matches. In 2002, Leonart Adelman and colleagues devised a “DNA computer” that could solve a logic problem with 20 variables:

Part B represents the “answer” to the problem, which DNA molecules can “find” faster than a silicon computer. DNA is particularly effective at problems allowing highly parallel computation.

In general, a DNA computation system requires three stages:

(1) Store the input data into DNA sequences, by constructing the DNA molecules.

(2) Mix the DNA molecules together, and let diffusion accomplish its task. Much is made of how this stage takes zero energy. Overall, though, one has to count the energy involved in DNA synthesis for step 1.

(3) Read the answer. Various ways to read the answer use, for instance, restriction enzymes that cleave specific double-stranded sequences, or series of electrophoretic gels that capture DNA strands with unpaired base sequences.

More recently, DNA computers are now making logic gates out of DNA, using molecules that combine information storage with enzyme capabilities, so-called DNAzymes. Such a computer was able to determine square roots of a number up to 15.

But how useful is such a “machine”?

One use is to store computing power in a kind of “instant package,” to answer a question in a remote location where electricity is unavailable. An example is a chain of “biochemical transistors” to test blood for the presence of the malaria parasite. Another example would be in devising nanobots to enter the human bloodstream for nanosurgery. The advantage here would be DNA’s high information density, compared to any known or projected silicon chip.

Will DNA computers ever be faster and better than silicon? If you have any thoughts on this, let us know.