DNA is an appealing archival medium because it is extraordinarily dense and requires no energy to maintain, but it degrades, and the codecs used to encode data into it cope poorly with the accumulated damage of long storage. A team spanning the School of Computing and Data Science and the Department of Electrical and Computer Engineering in the HKU Faculty of Engineering has approached the problem from the computation side rather than the chemistry side.
Their codec, Gungnir, applies a blockchain-style strategy: it generates candidate reconstructions of the original data and then verifies them, spending computing effort to recover information that is physically ambiguous. The approach tolerates error rates of up to 20 per cent, around four times better than existing methods. In extreme testing it achieved lossless recovery from DNA damaged to the level expected after roughly 400 years of storage.
The practical consequence is a change in the lifespan of a DNA archive, which the team puts at an extension from under a decade to several centuries. Because the error correction strengthens as available computing power grows, the researchers suggest that preservation over multiple millennia could eventually become feasible without any change to the storage medium itself. The work was published in Nature Communications and the code released as open source.
Trading computation against physical durability is an unusual design choice, and an attractive one for archives, where retrieval is rare and latency tolerable. It also implies that data written today becomes progressively more recoverable as hardware improves, inverting the usual relationship in which old media become harder to read. For institutional repositories facing decades-long retention mandates, that inversion is the most interesting claim in the paper.