Serving technology enthusiasts for more than 25 years. TechSpot is the place to go for tech advice and analysis you can trust.
Looking forward: Researchers from around the globe are adopting DNA-based storage. The combination of digital data with biology could be the answer to bridging the gap between the two, but there are still a few obstacles that slow down market and industry adoption.
For a few decades, DNA sequencing has been hailed by visionary solutions as the future of storage. The data encoding issue was solved by biology a few billions of years ago. We could learn from nature as we prepare to expand our digital world to 180 zettabytes (180 billion terabytes) by the year 2025.
Israeli Researchers say they have discovered a way to improve the data retrieval, which is currently one of the biggest challenges facing DNA storage technology. A team from Technion, Israel Institute of Technology, used a specially trained AI model to accelerate data recovery by 3,200. The process is still slower than the “modern” technologies on the market.
DNAformeris the AI tech in question. It is based on a model of a transformer trained by Technion scientists on synthetic data. Technion researchers also created the data simulator that fed DNAformer. The model can reconstruct accurate sequences of DNA from error-prone copies. It can also boost data integrity thanks to a custom algorithm that is designed to work with DNA.
DNAformer retrieves data much faster than any other method. The AI model is able to read 100 megabytes 3200 times faster than existing methods, and does so without losing data. The accuracy is also improved by “up to” 40%, which can further reduce the retrieval process time.
Israeli researchers tested DNAformer on a tiny data set of 3.1 megabytes, which included a color image, a 24 second audio clip, an article about DNA storage and some random data. The last data set was used to demonstrate how the model behaves when dealing with compressed or encrypted digital data. The official studystates that the team achieved “data rate” 1.6 bits per base (DNA) in a high noise regime. This reduced the time required to read back the data from several days to only 10 minutes.
According to the Technion team, DNAformer would be further developed for different data storage requirements. The technology is adaptable and can scale up to different scenarios. Researchers are already considering “market demands” as well as future improvements in DNA sequence to improve their AI technology.
