A DNA computer adds two 25-bit numbers by favoring the correct answer

A calculation can be encoded so that the most stable arrangement of molecules corresponds to the answer. In a paper published in Nature on September 16, researchers built a DNA computer that ran ten programs, including the addition of two 25-bit numbers. They call the addition a 100-bit computation because that count includes 50 input bits, 25 carry bits between positions and 25 output bits; it does not mean adding two 100-bit numbers.

The system uses a long DNA strand as a scaffold with positions to which short strands attach. The set of strands supplied determines the program and the input. Each piece also interacts with its neighbors: compatible combinations bind more stably, while poorly matched pieces can detach and be replaced. A sequence of molecular choices thus represents the logical steps of the calculation, and the favored assembly contains the answer.

Tristan Stérin, Abeer Eshra and colleagues tested tasks ranging from parity detection—determining whether eight binary digits contain an odd or even number of ones—to multiplication and division. They read outputs through fluorescent signals and compared them with control assemblies whose answers were known. In short systems they distinguished outputs after cooling procedures lasting about a minute; they also changed the inputs repeatedly in a simple program, with signals still distinguishable through 25 renewals.

The larger addition required 25 positions on the scaffold and took much longer: under some conditions the readout took about an hour, and a harder configuration reached roughly 14 hours. The team had to prepare the strands and heat and cool the mixture. The thermodynamic argument is that the answer was designed to be energetically favored during assembly; the study did not measure an electricity-use advantage over electronic computers.

The contribution is a reprogrammable platform showing that several kinds of calculation can emerge from reversible competition between molecules without prescribing a rigid sequence of steps for every piece. Practical applications would still require evidence about scale, speed and total energy cost. For now, the experiment opens a way to build chemical circuits, not a replacement for silicon processors.