Archival data storage using DNA proximity graphs
This paper introduces SPAN, a synthesis-free DNA archival storage method that encodes data as spatial patterns of randomly barcoded molecules on an optically patterned surface, offering a potentially five-orders-of-magnitude cost reduction in writing compared to existing DNA storage techniques.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
The world is drowning in data. Every day, sensors, cameras, and automated systems generate more information than humans can possibly read, let alone store. Most of this data is not meant to be looked at again; it is the digital equivalent of a receipt or a security camera feed that is recorded once and then forgotten. Yet, laws and history require that we keep these records safe for decades, or even centuries. The problem is that the current ways we store this information are expensive and fragile. Magnetic tapes and hard drives degrade over time, requiring constant replacement and power to keep them running. Scientists have long looked to DNA, the molecule of life, as a solution because it is incredibly dense and can last for thousands of years without electricity. However, writing data into DNA has been prohibitively expensive, as it traditionally requires a slow, chemical process to build specific sequences of genetic code for every single piece of information.
A team of researchers at the Massachusetts Institute of Technology has proposed a different way to use DNA for this deep storage, one that avoids the costly chemical writing process entirely. They call their method SPAN. Instead of carefully building a unique DNA sequence for every bit of data, they use a pre-made library of DNA molecules that are all identical except for a random, unique ID tag attached to each one. They then write the data not into the DNA itself, but onto a flat surface using light, creating a physical pattern of bumps and flat areas. When they pour the DNA library over this surface, the molecules stick to the flat areas but bounce off the bumps. The data is now hidden in the pattern of where the DNA landed and where it did not. To read the data back, the researchers use an enzyme to link the ID tags of DNA molecules that happened to land next to each other. By sequencing these linked pairs and mapping their connections, they can reconstruct the original pattern of bumps and flat spots, effectively recovering the stored information.
The researchers demonstrated this concept by writing a small image of the MIT logo onto a glass slide. They used a light-based technique to create a grid of tiny bumps, each 30 micrometers wide, arranged in a 17 by 8 pattern. They then applied their DNA library, which contained two types of molecules, each carrying a unique 25-letter genetic ID. An enzyme reaction was performed on the surface, which acted like a molecular glue, joining the IDs of any two DNA molecules that were sitting close to one another. After washing away the unlinked DNA, the researchers sequenced the remaining linked pairs. They fed this data into a computer program that mapped the positions of the molecules based on how often they appeared together. The result was a digital map showing where the DNA was dense and where it was missing. By comparing this map to the original grid, they successfully recovered 123 out of the 136 bits of information that made up the logo, a success rate of about 90 percent.
The significance of this work lies not just in the successful recovery of the image, but in the economics of the process. The most expensive part of traditional DNA storage is the writing phase, where chemicals are used to synthesize specific sequences. In this new method, the DNA library is synthesized once and can be used for any data, making the cost of writing data independent of the data itself. The researchers estimate that using their current experimental setup, the cost to write data is already four times lower than the most economical existing DNA synthesis methods. They project that if they use established, high-speed optical writing technologies similar to those used for Blu-ray discs, the cost could drop by another five orders of magnitude, becoming cheaper than flash memory for long-term storage.
However, this savings comes with a trade-off. While writing is cheap, reading the data is currently much more expensive than with standard DNA storage because the system requires sequencing a vast number of molecular connections to reconstruct the image. The researchers calculated that for data that is written once and rarely, if ever, read again, this trade-off is favorable. If the data is only accessed once in a hundred years, the low writing cost outweighs the high reading cost. They suggest that as sequencing technology continues to become cheaper over time, retrieving these archives will become even more affordable, making this a viable path for preserving the vast amounts of data that the modern world generates but rarely uses.
The success of the method depends on the physical relationship between the size of the written features, the spacing of the DNA molecules, and the distance over which the enzyme can link them. In their experiment, the DNA molecules were spaced apart, and the enzyme could link molecules up to a certain distance. This spacing was sufficient to clearly distinguish the 30-micrometer bumps from the flat areas. The researchers note that for this to work on a commercial scale, the spacing of the molecules and the reach of the enzyme would need to be scaled down to match the much smaller features of high-density optical writing, which are measured in nanometers. They believe this is possible, as the principles of molecular interaction do not change with scale, only the precision of the tools required.
This approach represents a shift in how we think about molecular storage. Instead of programming the information into the sequence of the molecule itself, the information is stored in the spatial arrangement of randomly assembled parts. The randomness of the DNA library is not a problem to be solved but a feature to be used; the unique ID tags allow the computer to figure out where every piece of the puzzle belongs based on its neighbors. This method turns the physical proximity of molecules into a data storage medium. While the current demonstration is small, encoding only 17 bytes of data, the principles are scalable. The researchers suggest that in the future, this could be extended to three-dimensional materials, such as gels, to increase storage density even further, or that the size and shape of the written features could be varied to store more than just a simple on or off signal at each location.
The study also highlights the potential for creating archives that are nearly impossible to tamper with. Because the data is distributed across millions of molecules and their connections, altering a single piece of information would be incredibly difficult without disrupting the entire pattern. This makes the system potentially ideal for legal or historical records that must remain unaltered for centuries. The researchers acknowledge that challenges remain, particularly in making the system random-access, so that a user can retrieve a specific file without having to reconstruct the entire map of the storage medium. They propose that adding specific "zip codes" to different sections of the DNA library could allow for selective retrieval, though this would add complexity to the writing process.
Ultimately, the paper presents a compelling argument for a new kind of archival storage. It suggests that by decoupling the writing process from the specific data being stored, we can drastically reduce the cost of preserving the world's information. The method relies on established technologies for writing patterns and reading DNA, combining them in a novel way to solve the problem of cost. While the reading cost is currently high, the trajectory of sequencing technology suggests it will fall, making this a promising solution for the deep future of data. The researchers have shown that it is possible to write information as a physical pattern and recover it from the relationships between random molecules, proving that digital information can be stored not just in the code of life, but in the space between its parts.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.