This is a system that allows the team to specify a piece of DNA with a desired shape and function, and then execute a molecular program to assemble it in a test tube. As an example, they used their system to construct a piece of DNA that walks along another strip of DNA. The researchers behind the approach predict that biochemical programming "languages" inspired by their work could let bioengineers create any number of desired molecular products and processes.
At the heart of the group's system are hairpin-shaped strands of DNA each about 10 nanometers long with three specific binding sites called "toeholds".
These hairpins can snap together in specific ways. When a hairpin is closed, for example, two out of its three binding sites are unavailable. But, if a suitable strand of DNA docks with it, the hairpin springs open.
A reaction between two DNA strands can also free up the exposed site on an attached hairpin, causing it to close once more.
In computer terms, the hairpins act as interconnected logic gates. "This elementary unit has one input port and two output ports," says Pierce. "And they can interact – the input port of one can receive an input from the output port of another."
The group has also developed a graphical way to represent the state of these molecular building blocks and the step-by-step interactions between them. These "reaction graphs" allow them to map out the assembly and disassembly steps needed to produce a desired sequence of reactions.
The necessary reactions are then translated into specific sequences of complementary DNA base pairs with the requisite binding characteristics. Finally, the program runs in a test tube that contains the specified mix of molecules.
Still, he says, a few years ago audience members laughed when he said he wanted to create a compiler to automate the process of encoding desired functions into DNA sequences. "Our field has now progressed to the point where the real question is not whether it can be done, but how far it can be pushed."
Some of Pierce's peers believe this kind of systematic biomolecular programming can be pushed very far indeed.
"It's great work," says computer scientist Erik Winfree, who is also at based Caltech, but was not involved with the work. "What's remarkable is that it develops a general way of creating a very diverse set of chemical reaction pathways. It opens a lot of doors."