Computers made from DNA have previously only been able to store information or perform computations on it – now a new device can do both.
A computer made from DNA that can solve basic chess and sudoku puzzles could one day, if scaled up, save vast amounts of energy over traditional computers when it comes to tasks like training artificial intelligence models.
 |
| A DNA computer can solve simplified chess and sudoku puzzles Kevin Lin |
DNA devices have a number of potential advantages, such as being able to safely store vast amounts of information, in microscopically tiny volumes, for millennia. One drawback of previous DNA systems is that they tend to focus on storing data in this way or on using it for computation, but doing both at the same time has proved difficult.
Now, Albert Keung at North Carolina State University and his colleagues have developed a computing system that uses cellulose microparticles to store many DNA strands at the same time, which makes it both more stable and reusable than previous systems. “We found that this marriage of DNA with a synthetic material gives you a whole host of new practical capabilities that weren’t possible before,” says Keung.
The DNA holders resemble a branched nerve cell, allowing them to store DNA-encoded data extremely densely (at 10000 terabytes per cubic centimetre) but also protect the DNA molecules, keeping them stable for long periods of time. DNA stored in this way at 4°C (39.2°F) should remain intact for 6000 years, says Keung.
 |
| A microscope image of the microparticles used to store DNA Prof. Orlin Velev group, NCSU |
Unlike other DNA computers, this set-up means that the DNA isn’t destroyed during computation. Instead, the team uses an enzyme to transcribe the DNA strands into RNA, which can then be read out by a sequencing machine, leaving the original data intact – a bit like making a copy of a document file before you start altering it. To demonstrate the system’s abilities, the team encoded three JPEG pictures – of a phosphorus symbol from the periodic table, a zebrafish embryo and a muscle cell – in around 2000 DNA strands and read each out 10 times.
Keung and his team also used the system to solve simple chess and sudoku problems, based on a 3-by-3 grid. They did this by loading each of the roughly 1000 possible board configurations into the DNA microparticles and then transcribing them into RNA. Then, they used an enzyme to eliminate any solution that would break the rules of chess or sudoku, leaving behind only valid solutions.
DNA computers are particularly suited to solving these kinds of problems, says Keung. If scaled up, they could offer advantages over expensive and energy-intensive processors of the kind currently used to train AI, he says. “Imagine… replacing much of that with low-cost, low-footprint, low-power, highly parallelised DNA-based molecular computation.”
Having very stable DNA particles while still being able to access and use them is impressive, says Matthew Patitz at the University of Arkansas, but the number of DNA strands that can be accessed and computed on will still need to be scaled up before the system becomes truly practical.
Journal reference:
Nature Nanotechnology DOI: 10.1038/s41565-024-01771-6