As global data generation outpaces the physical limits of traditional silicon and magnetic storage media, synthetic DNA has emerged as a compelling alternative. With its astonishing information density and multi-millennial durability, DNA could theoretically archive the world's data in a fraction of the space required by modern data centers.
However, a massive economic and operational barrier stands in the way: the synthesis bottleneck. While the cost of reading DNA (sequencing) has plummeted exponentially over the last two decades, the cost of writing DNA (de novo synthesis) remains prohibitively high. As reviewed by Yu Meng et al. in High-throughput DNA synthesis for data storage, writing digital data into DNA using standard base-by-base chemical synthesis is several orders of magnitude more expensive than magnetic tape, creating a multi-thousand-dollar overhead for even megabyte-scale datasets.
To make molecular storage commercially viable, researchers are shifting away from traditional, slow, and expensive nucleotide-by-nucleotide chemical synthesis. Instead, they are developing architectural, combinatorial, and algorithmic strategies designed to minimize or entirely bypass the overhead of custom DNA synthesis.
The Analogy: From Woodblock Printing to Movable Type
Traditional de novo DNA synthesis is akin to ancient woodblock printing. To store a specific file, a unique sequence of nucleotides (A, C, T, G) must be custom-manufactured from scratch. If the file changes or a new file is introduced, an entirely new molecular sequence must be synthesized. This "one-time-use" paradigm introduces severe financial and temporal overhead.
To break this bottleneck, recent architectures introduce a "movable type" paradigm. Instead of synthesizing custom sequences for every unique file, systems leverage pre-synthesized, reusable DNA building blocks.
Traditional De Novo Synthesis (Woodblock Printing):
[Custom File] ---> Synthesize Base-by-Base: A-T-C-G-G-C-T-A... (Slow & Expensive)
Combinatorial / Movable Type Storage:
[Pre-synthesized Blocks] ---> Assemble/Mix-and-Match ---> [Encoded Data] (Fast & Cheap)
For example, Chenyang Wang et al. developed a system detailed in Cost‐Effective DNA Storage System with DNA Movable Type that utilizes a library of pre-fabricated DNA fragments. By arranging these standardized "movable type" units into specific linear combinations, the authors successfully encoded and retrieved tens of kilobytes of data at a fraction of the cost of custom synthesis.
Similarly, Roman Sokolovskii et al. explored this concept in Coding Over Coupon Collector Channels for Combinatorial Motif-Based DNA Storage. Instead of relying on precise nucleotide sequences, information is encoded by the presence or absence of specific pre-synthesized metabolic motifs. By framing the retrieval of these motifs as a Coupon Collector problem, the authors demonstrated that data can be reliably reconstructed even from incomplete mixtures, drastically lowering both writing and sequencing overhead.
Bypassing Synthesis via Topological Modification
An even more radical departure from de novo synthesis involves using existing, native DNA molecules as a blank slate, recording data by altering the physical structure of the backbone rather than changing the sequence of bases.
In DNA punch cards for storing data on native DNA sequences via enzymatic nicking, Boya Wang introduced a "DNA punch card" system. Rather than building a strand base-by-base, the system takes a readily available, mass-produced DNA substrate (such as genomic DNA from a virus or bacterium) and uses a programmable enzyme (Pyrococcus furiosus Argonaute, or PfAgo) to introduce precise "nicks" or cuts in the phosphodiester backbone.
The presence or absence of a nick at a specific position acts as a binary 1 or 0. Because the underlying DNA sequence is never altered, this approach completely eliminates the synthesis bottleneck, offering a path toward high-speed, low-cost molecular writing. As noted by Olgica Milenkovic et al. in DNA-Based Data Storage Systems: A Review of Implementations and Code Constructions, such topological modifications allow researchers to treat DNA as a computable, rewritable substrate rather than a static, write-once medium.
Algorithmic Redundancy Reduction
When custom synthesis is unavoidable, the alternative strategy is to write less physical material. Traditional DNA storage pipelines rely heavily on physical redundancy—synthesizing multiple identical copies of a sequence—to safeguard against the high error rates (insertions, deletions, and substitutions) inherent to biochemical writing and reading. However, this redundancy multiplies synthesis costs.
To mitigate this, researchers are replacing physical redundancy with algorithmic intelligence. In Image processing in DNA, Chao Pan et al. introduced a "bits-for-brains" approach. Instead of using heavy error-correcting codes that require synthesizing extra DNA strands, the authors leveraged deep learning-based inpainting and quantization algorithms. When the DNA is read and found to be corrupted, the algorithm infers the missing data using structural priors, allowing high-quality image reconstruction from highly corrupted, low-redundancy DNA pools.
Furthermore, innovations in decoding allow systems to extract more value from messy, low-cost synthesis methods. Fast enzymatic synthesis methods often produce variable-length "runs" of identical bases, which are notoriously difficult to decode. In Coding for Optimized Writing Rate in DNA Storage, Siddharth Jain et al. designed coding schemes that exploit the statistical properties of these variable-length runs, maximizing the "writing rate" per unit of synthesis time. Complementing this, Seong-Joon Park et al. demonstrated in Reducing cost in DNA-based data storage by sequence analysis-aided soft information decoding of variable-length reads that utilizing the "soft information" (quality scores) from sequencers prevents the need to discard imperfect reads, driving down the required writing redundancy.
Architectural Paradigms: Random Access and Reusability
For DNA storage to compete with silicon-based infrastructure, it must move beyond sequential, archival "write-once, read-all" configurations. If a user must sequence an entire pool of DNA just to read a single file, the operational cost remains unsustainable.
Early breakthroughs by S. M. Hossein Tabatabaei Yazdi et al. in A Rewritable, Random-Access DNA-Based Storage System established that by designing mutually uncorrelated address strings to flank data blocks, specific files could be targeted and amplified via PCR, enabling random access and in-place editing. Building on portable sequencing platforms, S. M. Hossein Tabatabaei Yazdi et al. later demonstrated in Portable and Error-Free DNA-Based Data Storage that combining these address constraints with iterative alignment pipelines allows error-free data retrieval even on low-cost, handheld nanopore devices.
Random Access Architecture:
[Pool of DNA Strands]
├── [Address A]--[Data Block 1]--[Address A'] ---> PCR Amplifies Only File 1
└── [Address B]--[Data Block 2]--[Address B']
More recently, the focus has shifted toward mimicking conventional hardware architectures like Flash memory and RAM to amortize the initial synthesis cost over many cycles of use. In DNA Tails for Molecular Flash Memory, Chao Pan et al. introduced an architecture that appends variable-length single-stranded "tails" to existing double-stranded DNA molecules, allowing data to be updated and read dynamically without re-synthesizing the core substrate.
To bridge the gap to functional filesystems, Billy T. Lau et al. developed MDRAM (Magnetic DNA-based Random Access Memory), presented in Magnetic DNA random access memory with nanopore readouts and exponentially-scaled combinatorial addressing. By anchoring DNA strands to magnetic beads and employing a combinatorial addressing scheme, the authors created a reusable proto-filesystem capable of managing millions of files, achieving a 10x reduction in reading costs and decoupling the high upfront cost of synthesis from daily data management operations.
Key Foundational Work
Acetylated and Methylated β-Cyclodextrins as Viable Soluble Supports for the Synthesis of Short 2′-Oligodeoxyribo-nucleotides in Solution (2012) — contributed 2 concept elements to this direction.
The completion of the Mammalian Gene Collection (MGC) (2009) — contributed 1 concept elements to this direction.