Rewriting existing DNA rather than building from scratch
Somewhere between the ancient code of life and the modern hunger for data, researchers at Peking University have found an unexpected bridge. Rather than constructing DNA from nothing, they have learned to rewrite what already exists — borrowing a mechanism cells use to adapt across a lifetime — and in doing so, they have made the dream of biological data storage measurably more real. The question humanity has long carried, of how to preserve its knowledge without consuming the planet to do so, now has a quieter, more elegant candidate for an answer.
- The world's data centers are devouring electricity at a pace that cannot hold, and the pressure to find a radically different storage medium has never been more urgent.
- Traditional DNA data storage has been strangled by its own complexity — hand-carving every letter into stone, one expensive, error-prone synthesis at a time.
- The Peking University team short-circuited that bottleneck by using methylation, a natural cellular process, to rewrite existing DNA strands rather than build new ones from scratch.
- Sixty volunteers with no scientific training successfully encoded text using iDNAdrive, an app the researchers built to pull this technology out of the laboratory and into ordinary hands.
- A startup is now claiming it can store all the world's data in a small box within three years — and for the first time, that claim does not sound entirely like science fiction.
For years, the promise of DNA as a data storage medium has been tantalizing and frustrating in equal measure. A single gram could theoretically hold roughly 215,000 terabytes — yet the process of writing data into DNA has remained slow, expensive, and unforgiving, requiring the construction of entirely new genetic sequences from raw components.
A team from Peking University and three partner institutions has found a way around that wall. Instead of building new DNA strands, they rewrite existing ones using methylation — a natural epigenetic process that living cells use to adapt during their lifetimes. By engineering 700 DNA components that act like movable type, the researchers can stamp data onto DNA already present, encoding information into what they call "epi-bits." They demonstrated the method by encoding a Chinese tiger rubbing and a panda photograph, working at roughly 350 bits per reaction — slow by conventional standards, but a meaningful leap forward, and far cheaper because no new DNA needs to be synthesized.
The team then did something unusual for a research group: they built a consumer-facing app. iDNAdrive allowed sixty volunteers with no biotechnology background to encode text into DNA themselves, democratizing a technology that had previously lived behind locked laboratory doors.
The timing carries weight. Data centers are consuming electricity at a rate the world can ill afford to sustain, and at least one startup is claiming that a fully autonomous DNA storage system could hold all of humanity's data in a small box within three years. Whether or not that timeline holds, the field has visibly shifted — from isolated experiments toward something that begins to resemble a real, scalable system.
For years, scientists have known that DNA could theoretically hold staggering amounts of information—a single gram could store roughly 215,000 terabytes. The problem has never been the medium itself, but the exhausting, expensive, error-prone process of getting data into it. Building custom DNA sequences from scratch, what researchers call de novo synthesis, requires painstaking labor and specialized equipment. It's like trying to write a book by hand-carving each letter into stone.
A team from Peking University and three collaborating institutions has found a different path. Rather than constructing new DNA strands, they're rewriting existing ones using a natural biological process called methylation. This is an epigenetic modification that happens during an organism's lifetime—not passed down through generations, but occurring in living cells. The researchers realized they could use this mechanism to essentially print data onto DNA that's already there, transforming the entire workflow.
The breakthrough required engineering 700 different DNA components from nucleic acids, each acting like a piece of movable type. By methylating specific locations on existing DNA strands, the team can encode information into what they call "epi-bits." The analogy is apt: instead of chiseling every word into stone, you're using a word processor. They tested the method by encoding images—a Chinese tiger rubbing containing 16,833 bits of data, a panda photograph with 252,504 bits—at a rate of roughly 350 bits per reaction. It's glacially slow compared to even the oldest hard drives, yet it represents a dramatic leap forward from conventional synthesis. More importantly, because the technique reuses existing DNA rather than creating new strands, the cost drops substantially.
The researchers didn't stop at the laboratory. They built a software application called iDNAdrive designed for people without biotechnology training. Sixty volunteers with no background in the field successfully used the app to manually encode around 5,000 bits of text data. The software makes DNA storage accessible beyond the walls of research institutions, democratizing a technology that once seemed locked behind specialized expertise. This shift from laboratory curiosity to user-friendly tool marks a genuine inflection point.
The timing matters. Data centers worldwide consume enormous amounts of electricity, and that appetite only grows. A DNA storage startup has already claimed that an autonomous system could theoretically hold all the world's data in a small box—and the company says such a device will be ready within three years. If that vision becomes reality at a price point that makes economic sense, it could fundamentally reshape how we store information at scale. The researchers themselves framed their work as pointing toward "parallel molecular information storage with prefabricated modularity," a way of saying the field is moving from one-off experiments toward systems that could actually work in the real world. The race to make DNA storage commercially viable has entered a new phase.
Citas Notables
The epi-bit framework demonstrates potential directions in parallel molecular information storage with prefabricated modularity— Peking University research team
La Conversación del Hearth Otra perspectiva de la historia
Why does the speed matter if DNA storage is supposed to be about density, not performance?
Speed matters because it determines whether the technology can actually be deployed. If encoding data takes weeks, nobody will use it, no matter how much information fits. The methylation approach is still slow by computing standards, but it's fast enough that the process becomes feasible.
So this methylation technique—is it something that only works in a lab, or could it scale?
That's the real question. The fact that they built iDNAdrive and had untrained volunteers successfully use it suggests it could scale. But there's a difference between sixty people encoding kilobits and a data center encoding petabytes. The infrastructure doesn't exist yet.
What makes this cheaper than the old way?
You're not synthesizing new DNA from scratch every time. You're reusing what's already there and just modifying it. That's like the difference between printing on blank paper versus printing on paper that's already been used. The raw materials cost less, and the process is simpler.
Is there a catch? Something that makes this sound too good to be true?
The speed is still slow. And we don't know yet how many times you can reuse the same DNA strand before errors accumulate. The lab results look promising, but real-world deployment at scale is still theoretical. That's why the startup claiming they'll have a commercial system in three years is worth watching—they'll either prove it works or they won't.
Why should anyone care about this if data centers work fine?
Data centers don't work fine—they consume massive amounts of electricity and keep growing. If DNA storage could replace even a fraction of that infrastructure, it would reshape how the world stores information and dramatically reduce energy consumption. That's not incremental. That's foundational.