Scientists Built Tiny Robots Out of DNA That Can Hunt Viruses and Deliver Drugs

A machine so small it could seek out a cancer cell, deliver a drug directly to it, or grab a virus and block it from infecting healthy tissue sounds like science fiction. But what if that machine were built not from metal or plastic, but from the same molecule that carries the blueprint for life? That’s the promise of DNA-based machines, a field that has quietly grown from a niche academic curiosity into one of the most ambitious areas in science and engineering.

A recent review paper published in the journal SmartBot traces the evolution of these molecular machines, from the simplest DNA structures dreamed up in the 1980s to today’s tiny robots capable of detecting viruses, delivering drugs, and even performing basic computations. Authored by researchers at Peking University, Stanford University, and King’s College London, the paper maps out where this technology is headed and the obstacles still in the way.

At its core, the paper argues that the future of DNA machines depends not just on biologists and chemists, but on mechanical engineers, computer scientists, and artificial intelligence. Building a working robot at the molecular scale demands the same kind of thinking that goes into designing a car engine or a factory robot arm, just a billion times smaller.

How DNA Robots Went From Smiley Faces to Virus Catchers

In the 1980s, a scientist named Nadrian Seeman first proposed the radical idea of using DNA not as a carrier of genetic information but as a building material. He recognized that DNA’s famous double-helix structure, combined with the predictable way its four chemical “letters” pair up, made it an ideal construction material at the molecular scale. For years, Seeman’s lab was one of the few in the world pursuing this idea, partly because making synthetic DNA was still expensive and difficult.

A major breakthrough came in 2006, when Paul Rothemund introduced what’s now called the DNA origami technique. The concept is deceptively simple: take a long single strand of DNA and fold it into a desired shape using hundreds of shorter “staple” strands that act like molecular paperclips. Rothemund demonstrated the method by creating tiny smiley faces and stars visible only under powerful microscopes.

From those flat smiley faces, the field rapidly graduated to three-dimensional structures: cubes, vases, gear-like shapes, and even a wireframe rabbit. But static sculptures were only the beginning. Researchers soon began asking a far more ambitious question: could these structures actually move?

DNA Robots With Joints, Hinges, and Moving Parts

Scientists have borrowed directly from the playbook of mechanical engineering to make DNA structures that move with purpose. In the everyday world, machines rely on joints: hinges that rotate, sliders that push back and forth, and linkages that convert one type of motion into another. At the DNA scale, researchers have figured out how to create molecular versions of all these parts.

Key to that feat are the physical properties of DNA itself. Double-stranded DNA, the classic twisted-ladder form, behaves like a stiff rod when it’s short enough, staying straight over lengths of about 50 nanometers. Single-stranded DNA, by contrast, is floppy. By combining stiff double-stranded segments as structural beams with floppy single-stranded segments as flexible joints, engineers can build molecular mechanisms that mimic real-world machines.

Researchers have assembled these building blocks into systems that allow rotation, sliding, and movement in multiple directions. One team built a human-shaped DNA robot with movable limbs.

But the paper is candid about the limitations. Shrinking engineering principles down to the molecular scale “is not a direct scaling‐down process,” the authors write. DNA joints are constantly buffeted by the random jiggling of surrounding molecules, which creates “positional jitter” and makes precise control far harder than with everyday machines. As these devices grow more involved, the accumulated wobbliness becomes a serious engineering challenge the authors call “structural floppiness.”

Powering and Programming DNA Robots

Building a tiny machine is one thing. Making it do something useful requires a power source, and this is where the field gets especially creative.

Electric fields can push and pull DNA structures because DNA carries a natural negative electrical charge. Magnetic nanoparticles can be attached to DNA machines, allowing researchers to steer them with external magnets — a particularly appealing approach for medical uses deep inside the body. Light and heat can trigger DNA strands to zip or unzip, causing the structure to change shape.

Perhaps the most elegant approach is called strand displacement. New DNA strands are introduced into a solution, where they compete with existing strands for binding partners. By carefully designing the sequences, researchers can make specific joints open or close in a set order, giving them precise control over multiple moving parts at once. The authors note that optimized strand displacement reactions “typically complete within minutes,” making this method practical for machines that need to reconfigure on the fly.

Each approach has trade-offs. Electric fields offer speed but lack fine control over individual joints. Magnetic steering can reach deep into tissue without invasive procedures, making it an appealing approach for medical applications inside the body, though it requires attaching extra materials to the DNA structure. Strand displacement offers the highest programmability but uses up its fuel strands in a single reaction, generating chemical waste. Future systems, the authors suggest, will likely combine multiple approaches, such as using rapid physical fields for speed alongside strand displacement for precision, to create truly self-directed molecular robots.

Designing these machines also requires serious computing power. Early DNA structures were painstakingly designed by hand, with researchers manually planning how each strand would route through the structure. Software platforms eventually automated much of this process, letting researchers import three-dimensional shapes and automatically generate the DNA sequences needed to build them.

One platform called MagicDNA marks a shift from shape-based design to motion-based design. Instead of simply asking “What does this structure look like?” the tool helps researchers ask “How does this structure move?” That distinction is critical for building functional machines rather than static sculptures.

What DNA Robots Can Already Do

DNA-based machines have been developed for targeted drug delivery, where a molecular container opens only when it encounters specific disease markers on a cell. Virus-capturing grippers have been designed that can physically grab viral particles and may interfere with their ability to infect cells. DNA walkers — molecular robots that take deliberate steps along a predefined track — have been developed as a platform for transporting molecular cargo.

One production advantage sets this field apart from conventional manufacturing. Because the assembly reactions use very dilute solutions, a single experiment can produce hundreds of millions to billions of identical structures or machines in one batch. That kind of output is essential for any technology that hopes to move beyond the lab.

Still, the gap between laboratory demonstrations and real-world use remains wide. The authors identify several hurdles: making DNA structures more durable in biological environments, scaling up production to industrial levels, and developing better ways to predict how these machines will behave. Artificial intelligence, they argue, will play a growing role in tackling these problems, from designing optimal DNA sequences to predicting mechanical behavior to automating the entire pipeline from concept to finished product.

What this review ultimately makes clear is that DNA machines occupy a genuinely new category; not smaller versions of existing technology, but something different in kind. They assemble themselves from the bottom up, operate in the thermal chaos of a living cell, and can be programmed with the same digital logic as a computer.

A field that started with a scientist folding DNA into smiley faces has arrived at molecular robots with moving joints, programmable logic, and the ability to interact with living systems. Whether they eventually rival the precision of life’s own molecular motors remains to be seen. Still, the engineering case for trying has never been stronger.