A DNA computer in a water droplet reaches its answer by settling into its lowest-energy state

Start with the tube. Short DNA strands and a much longer DNA scaffold go into a small amount of salt water, which is heated and then cooled ^[2]. The sequences decide which strand can attach to which, and to which position on the scaffold, and choosing those binding rules is the programming step ^[3]. "The clever part is that the binding process is competitive: the DNA molecules compete with each other to select a winner, which succeeds in binding to the scaffold; all of the jostling and competition process information and execute a computation," Damien Woods, a professor of computer science at Maynooth University in Ireland and a co-author of the study, told Live Science in an email ^[4]. "Eventually, the system settles down into its energetically-preferred state which encodes the answer to the computation," he said ^[5].

"A small droplet of liquid contains billions, and sometimes trillions, of DNA strands," Abeer Eshra, an assistant professor of computer science and a co-author, said in a statement ^[14]. Those strands are not each running a program. They compete for places on one scaffold, and the structure they settle into carries one answer ^[4]. Small calculations finished in under a minute even though the computation has to proceed through chemical reactions that take the same amount of time or longer ^[13]. More than 700 computations across ten programs also works out to roughly 70 runs per program ^[20]. That is a lot of repeats for a first demonstration.

As reported, the efficiency claim is a claim about design. The system is arranged so that the energetically favorable state is the correct answer, so no energy goes into forcing a calculation through a series of processing steps ^[6]. Live Science reports the scientists describing this as more efficient than the conventional computers in wide use today ^[8]; the comparison the study itself draws is with biological computers that wire living cells to traditional hardware ^[7]. No measured energy per operation is reported ^[21].

"They're trivial calculations you could easily do faster yourself, and a silicon computer would finish in an instant," Constantine Evans, a senior research fellow at Maynooth University and a co-author, told Live Science ^[18]. In my view that closes the speed question and leaves the open one exactly where the authors put it, on what a competitive-binding answer costs to reach ^[6].

Eshra pointed at narrower uses than general computing. "Molecular computers like this are not trying to replace electronic ones, but they could be used in biological environments, smart materials and archival DNA data storage," he told Live Science via email ^[15]. "Our work is a new direction for DNA data storage, since any data stored in such a system would have natural built-in error correction properties," he said ^[16].