AI can rapidly decode digital data stored in DNA in a matter of minutes instead of days

Artificial intelligence can read data stored in DNA strands within 10 minutes, not the day required for previous methods, and DNA storage approaches practical use of computing.

“DNA stores huge amounts of data in a very compact form and remains intact for thousands of years.” Daniella Bar-Lev At the University of California, San Diego. “In addition, DNA is naturally replicable and offers a unique advantage for long-term data storage.”

However, getting encoded information within DNA is a monumental challenge as the strands become confused when mixed and stored. During the data encoding process, individual strands may be replicated incompletely, and some fragments may be lost completely. As a result, data readings stored in DNA are similar to reconstructing books from boxes filled with pages that go beyond the finest, varied.

“The traditional methods suffer from this confusion and require several days of processing,” says Bar-Lev. The new approach “rationalizes this with AI trained to find patterns of noise,” she says.

Bar-Lev and her colleagues developed an AI-driven method called DNAFormer. The system includes a deep learning AI model trained to reconstruct DNA sequences, another computer algorithm to identify and correct errors, and a third decoding that brings everything back to digital data while correcting any remaining mistakes. Includes algorithms.

In experiments, DNAFormer can read 100 megabytes of DNA storage data 90 times faster than the next fastest method developed with traditional rules-based computing algorithms, while achieving better or equivalent accuracy. I did. The decoded data included colored images of the test tube, a 24-second audio clip from astronaut Neil Armstrong's famous Moonlanding speech, and text on why DNA is a promising data storage medium .

The team says they plan to develop a version of DNAFormer tailored to new technologies for encoding data into DNA. Omar Savery Technion – Israel Institute of Technology.

“Crucially, our approach doesn't rely specifically on us. [DNA] Synthesis or sequencing methods can be adapted to future, still undeveloped technologies that are more commercially viable,” he says.