Making DNA strands compute
Alan Turing once imagined that logical symbol-manipulating devices could lead to general-purpose computers, but he never imagined they could one day allow objects like our smartphones. Thanks to their theoretical studies and experiments, Damien Woods, David Doty and Erik Winfree have done the same with DNA strands and their results look promising.
The researchers managed to run 21 simple computing programs thanks to DNA strands that compute on 6-bit input data. “When we first ordered the DNA strands for the full system, we had only designed 3 programs for it” explains David Doty.
But once we started using the system, we realized just how much potential it has. By the end, we had designed and run a total of 21 circuits.
Choosing different sets of 100 tiles from the “library” of 355 tiles leads to many different structures that are like logical gates aligned in a row and can then produce results exactly like the electronic logical gates one can find in our modern computers.
For a specific 6-bit input, the simplest DNA nano-apps can copy, sort, recognize palindromes (i.e. reads the same backward as forward) or multiples of 3. But the researchers were also able to build more sophisticated programs like counting to 63, generating patterns, obtaining an unbiased choice from a biased random source, electing a leader or simulating cellular automata.
The DNA strands they used are not exactly the same ones you carry in your cells (they are not the simple DNA double helix structure that carries genetic instructions): here the DNA strands can be thought of as strings of 42 DNA bases (A, C, G or T) arranged in 4 domains of 10 or 11 bases, with each domain being the side of a square tile. The researchers built a “library” of 355 different DNA tiles which interact differently with one another: a first tile may stick to another one (through a “sticking bond” between several pairs of bases: A sticks to T and G sticks to C) while it won’t stick with a third one because it lacks a potential “sticking site”. Finally, they designed a whole set of these tiles from which you can choose a subset (call it the “program”) that will build a specific pattern by carrying out logical rules to decide which tiles stick next as the structure grows.
But to implement the process and compute with it – i.e. to “run” the program – two more components are needed. The first one is a barrel-shaped seed, called a DNA origami, of length 140 nanometers. The end of this structure only allows specific DNA strands to stick onto it. And these specific strands are the second component: the 6-bit input data.
Here an extra implementation detail is needed: in order to visualize DNA tiles as “coding bricks” that carry binary information, at the end of the experiment protein “labels” are added that attach to those tiles that encode a “1” bit so that they are easier to recognize. The researchers made their proof of concept by building circuits that operate on 6 bits of data at a time i.e. groups of 6 tiles each carrying or not carrying the protein to encode binary strings like 100101.
Programming at the lab bench
When all of these ingredients for the recipe are ready, the very last step is to mix, in a tube filled with salted water, the DNA origami, the tiles that will encode the input data and the tiles that will form the “computed circuit”: everything will self-assemble from one end of the DNA origami like a knitted scarf!
“After 1 day, the test tube is filled with billions of self-assembled scarfs and the computation is done. Then, we add protein labels and by using an Atomic Force Microscope (AFM) one can see where labels are present (meaning “1”) or not (meaning “0”) on the self-assembled scarf and can interpret the 6-bit output of the circuit” explains Damien Woods.
Our results are a proof of concept that DNA tiles can be used as a computing device since it has also shown that it is reliable: it has an overall per-tile error rate of less than 1 in 3,000, adds Erik Winfree.
Both Damien Woods and David Doty were theoretical computer scientists when beginning this research and they had to learn a new set of lab skills that are typically more in the wheelhouse of bioengineers and biophysicists. "When engineering requires crossing disciplines, there is a significant barrier to entry" explains Erik Winfree. "Computer engineering overcame this barrier and today's programmers do not need to know transistor physics. Our goal in this work was to show that molecular systems similarly can be programmed at a high level: unlike previous experiments on molecules specially designed to execute a single computation, reprogramming our system to solve different problems was as simple as choosing different test tubes to mix together" he adds.
Although DNA computers have the potential to perform more complex computations than the ones the researchers featured in their Nature paper, Damien Woods cautions that the point of this research is not to replace standard silicon microchip computers. "These are rudimentary computations and there is much work to do to allow computations on more than 6-bits of data or to have a room-temperature process with the same good overall per-tile error rate. But these results teach us more about how simple molecular processes like self-assembly can encode information and carry out algorithms. Biology is proof that chemistry is inherently information-based and can store information that can direct algorithmic behavior at the molecular level".
An international interdisciplinary team
Damien Woods is the former head of the Inria TAPDANCE team (2016-2018) and he is now a Professor at Maynooth University in Ireland where he also completed his PhD in 2005. His group’s work is funded by a European Research Council (ERC) consolidator grant (2018-2023) on molecular computing. With David Doty (University of California, Davis, USA) and Erik Winfree in Caltech (California Institute of Technology, Pasadena, USA), he worked for years on self-assembly of DNA strands and how it can be used to design computing systems. For these results - published in early 2019 in Nature - the experiments were done in Caltech (USA) and the analyses continued at Inria Paris and UC Davis.
"Diverse and robust molecular algorithms using reprogrammable DNA self-assembly", Woods, D., Doty, D., Myhrvold, C. Hui, J., Zhou F., Yin P., Winfree, E. Nature. volume 567, pages 366–372 (2019)