Back

A reconfigurable DNA memory architecture for hierarchical data management via programmable phase transitions

Ye, J.; Liu, A.; Wang, F.; Zhang, H.; Zhang, H.; Fan, C.

2026-01-18 synthetic biology
10.64898/2026.01.17.700124 bioRxiv
Show abstract

DNA has emerged as a promising medium for the post-silicon era of information storage due to its ultrahigh density and longevity. However, current systems are bifurcated, with solid-state systems providing robust cold archival but lacking accessibility, while fluidic molecular computing systems offer dynamic processing but suffer from low density and instability. This mutual exclusivity has hindered the development of hierarchical memory, a standard in modern computing, within molecular storage systems. Here, we bridge this gap by engineering a reconfigurable DNA memory architecture driven by programmable liquid-liquid phase separation (LLPS). Our system leverages sequence-based encoding to achieve an ultrahigh storage density of 7x1010 GB/g, approaching the theoretical limits of DNA accessibility. In its fluidic hot state, DNA droplets enable rapid data loading ([~]83.8% in 5 min) and function as an in-memory editing platform supporting versatile, addressable bit-level operations including selective erasure ([~]65.1%) and high-efficiency rewriting and replacement (>99%) via programmable strand displacement. Importantly, to resolve the stability trade-off, we engineered a programmable phase transition whereby the triggered assembly of a rigid tetrahedral DNA framework (TDF) armor transforms liquid condensates into robust armored droplets. This cold state confers exceptional resistance to enzymatic and physical degradation, projecting multi-millennial data stability. By enabling reversible transitions between a editable, high-density computing mode and a stabilized archival mode, this work establishes the architectural foundation for scalable molecular information storage capable of hierarchical data management.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.