DNA has formed the backbone of self-assembling logic circuits designed by a team at Duke University. To communicate, the circuits employ light-emitting molecules already widely used by biologists in their own experiments.
Chris Dwyer, assistant professor of electrical and computer engineering at Duke’s Pratt School of Engineering, said the technique could be used to build intelligent but tiny biosensors as well as nanoscale encryption keys.
Dwyer said the logic is form of diode-diode logic, one of the earliest approaches to digital computation used in electronics. Although it cannot form all the possible Boolean logic gates, it can be used to build simple computers from AND and OR gates. In the Duke University scheme, the diodes of an electronic circuit are replaced with chromophores - light-absorbing elements - attached to segments of DNA.
DNA-linked chromophores, particularly those that fluoresce, are used widely in biological experiments as they make it easy to identify the locations of genetic elements within a cell. Theodor Förster found in 1948 that chromophores can pass energy to other, different chromophores close by through a coupling process. Biologists often use this in Förster or fluorescence resonance energy transfer (FRET) to show when molecules such as proteins are coupled together in complexes.
In FRET experiments, the two chromophores used emit different wavelengths of light but only one is excited by a light source. Only when the two molecules are so close as to allow an energy transfer will the other chromophore light up. The logic circuits act as biological detectors by placing analyte receptors at a gate to bind to chemicals in a sample and, if they are present, disrupt the normal operation of the gate.
The Duke University researchers use a much more complex array of chromophores to pass signals to each other and, as a result, build logic gates. “The difference is that we chain them together,” said Dwyer.
“With diode-diode logic, you tie the outputs of two diodes together. Depending on the impedance you can make it function like an AND or an OR gate. Each donor-acceptor chromophore pairs that we can pattern is like a diode. If you share a common acceptor between two donors, it looks like an array of diodes tied to a single node. There is a really strong analogy to electrical circuits,” Dwyer explained.
“We can tune the gate’s behaviour by changing the positions on the grid to make the coupling efficiency greater or smaller. This gives us tunable ‘resistance’ for these diodes pairs. You can require all of them to be excited to give an output, which is used as an AND gate,” he added. “And if you bring them closer together any one can light up to produce an output, like an OR gate.”
Using nine different types of DNA fragment, the logic circuits designed by the Duke team form into grids that look like waffles in atomic-force microscopy images. Each individual waffle is 80nm across.
“The self-assembly process doesn’t require a patterned substrate. The sequences are designed such that they tie themselves together into a grid shape,” Dwyer explained. “It’s like taking pieces of a puzzle, throwing them in a box and, as you shake the box, the pieces gradually find their neighbours to form the puzzle.”
The most difficult part is designing the sequences to self-assemble, said Dwyer. Initial efforts to build arbitrary shapes from DNA focused on regular grids. To produce practical logic circuits, the sequences need to be less regular. “The structures we use now have complex addressability so that we can put down any molecule at any position on that grid.
“The new piece of the puzzle is that we can very precisely control the relationship between the donors and the acceptors. I don’t know any other way to get that kind of control,” Dwyer claimed. “This is work that has been building for 30 years.”
Dwyer said the self-assembly process, as with other DNA computer techniques, is error prone. But he argued that is it possible to focus on structures that deliver good results and that, in the future, sensors based on the chromophore-DNA logic could probably cope with 30 per cent error rates in construction.
“What this does is let you pack a lot of sensors into a diffraction-limited spot, just hundreds of nanometres across,” said Dwyer. By using this logic, many light-emitting sensors could fit into that space compared with conventional techniques that could only discriminate a few. “This lets you multiplex a bunch of things within that area, so it could improve the density of gene arrays.”
By performing logical operations withint the sensor, Dwyer said it should be possible to speed up diagnoses that today have to be done in a lab when multiple samples have to be analysed. “At a pharmacy or local level, that would have a dramatic impact in changing the business of healthcare. We are basically building nanoparticles that are 80nm on a side with a little computation inside that could impact a whole range of applications.”
“The real key is to bring in more sophisticated molecular beacons,” added Dwyer. With conventional FRET experiments, it is one chromophore for each analyte so that with ten different molecules, the maximum number that can be counted is ten. “We can put a binary encoding of those channels and instead of getting ten we get 2¹⁰ channels. We want to use the combinatorics of Boolean logic to conserve chromophores.”
Dwyer said applications could go beyond medicine. “One of the things we have thought of using it for is in the area of physically unclonable functions, to make a physically secure encryption key. It is basically impossible even with the best analytical chemical tools to discern the structure of chromophores within the array.”
Designing the structures today requires intense computation. “It is heavily computationally bound. It took 500 machines running 24/7 for a couple of weeks to solve a very small subset of the problem.”
Although the chromophore logic makes it possible to use a smaller number of chromophores, Dwyer said he would like to see more available to extend the computational range of these devices. Out of 600 known chromophores only about 40 are useful in biological work and, of those, only 20 are suitable for use in this type of logic circuit.
Up to now, there has not been a lot of demand for different chromophores. Although heavily protected by patents, not many researchers go beyond GFP and its relatives for FRET and similar experiments. And, as it’s not yet possible to design a chromophore on a computer - the work to create them is largely empirical - the research is expensive. However, if applications such as this one create a demand, more synthetic chemists may choose to work on an expanded library of chromophores.