The Sequential Access Bottleneck
DNA-based data storage offers unprecedented information density, but early architectures suffered from a fundamental limitation: they functioned like digital magnetic tape. Because billions of DNA strands float together in a well-mixed molecular "soup," retrieving a specific file traditionally required sequencing the entire pool. As noted by Meng et al. and Tabatabaei Yazdi et al., this full-pool sequencing introduces massive latency, high computational overhead, and unsustainable costs, rendering DNA impractical for active, searchable databases.
To transition molecular storage from a static, write-once archive into a functional database, researchers must implement random access—the ability to isolate and read a specific data block without reading the rest of the library. Gao et al. point out that while payload optimization is highly advanced, a practical storage system requires integrated indexing and header mechanisms to uniquely identify individual strands within a complex biochemical mixture.
Sequence-Level Addressing via Cross-Bifix-Free Codes
The most common mechanism for selective retrieval relies on Polymerase Chain Reaction (PCR). By appending short, unique "address" sequences (primers) to the boundaries of data payloads, researchers can introduce matching physical primers to amplify only the desired strands, physically isolating them from the background pool.
However, DNA strands are prone to non-specific binding, or cross-hybridization. If a primer sequence accidentally binds to a data payload or a different file's address, the wrong data will be amplified, corrupting the readout.
Staggered Cross-Hybridization (Bad):
Address A: [---Prefix---][---Suffix---]
Address B: [---Prefix---][---Suffix---]
^^^^^^^^^^^^
Dangerous overlapping region
To prevent this, Tabatabaei Yazdi et al. introduced the use of mathematically constrained, mutually uncorrelated or cross-bifix-free address strings. A set of sequences is cross-bifix-free if no prefix of one sequence matches the suffix of any other sequence in the library. This geometric constraint prevents strands from shifting and binding to one another in a staggered configuration during the PCR annealing phase.
By enforcing this rule, the authors ensured that primers bind with absolute specificity. Building on this foundation, Kiah et al. developed explicit mathematical constructions for these primer codes using cyclic and reversible cyclic codes, bypassing the need for computationally heavy heuristic searches. Using these optimized addresses, Tabatabaei Yazdi et al. demonstrated error-free random access and data recovery even when using low-cost, high-error portable nanopore sequencers.
Topological Addressing on the Molecular Backbone
While sequence-level addressing solves the retrieval problem, it still requires the costly de novo synthesis of custom DNA strands for every new piece of data. An alternative paradigm, reviewed by Milenkovic et al., is topological backbone encoding. Instead of writing data into the sequence of $A, C, G,$ and $T$ bases, information is written by introducing site-specific enzymatic "nicks" (controlled breaks in the sugar-phosphate backbone) into existing, naturally occurring DNA molecules.
Topological Nicking (DNA Punch Card):
Strand 1: ===================== (No nicks = Bit 0)
Strand 2: ======= === ======== (Nicks introduced = Bit 1)
Kasra Tabatabaei et al. demonstrated that this approach enables single-bit random access without sequencing. By introducing two adjacent nicks on a strand, the small intervening segment dissociates, leaving a short, exposed single-stranded pocket. This pocket acts as a highly localized physical address. Instead of running a PCR reaction, researchers can introduce fluorescently labeled probes that hybridize exclusively to these exposed regions. The data can then be read optically, transforming native DNA into a high-speed, sequence-free "punch card."
This topological strategy also solves the problem of rewritability. Because data is decoupled from the primary sequence, Pan et al. showed that these nicks can act as a rewritable metadata layer. Administrative data, such as ownership or timestamps, can be erased by sealing the nicks with ligase enzymes and re-patterned with new cuts, avoiding the need to re-synthesize the underlying molecular strands.
Physical Substrates and Reconfigurable Media
In a standard liquid-phase random access system, performing a PCR readout consumes primers and risks depleting or cross-contaminating the master data pool over time. To make molecular random access truly sustainable, researchers have turned to physical substrate engineering and reconfigurable chemical modifications.
- Immobilized Substrates: Lau et al. developed Magnetic DNA Random Access Memory (MDRAM), where DNA templates are physically anchored to magnetic beads. During a random access query, the target file is amplified via PCR, and the magnetic substrate allows the master library to be pulled aside and preserved while the amplified product is washed off for sequencing. This prevents the degradation of the original archive across repeated reads.
- Metastable Hybridization: Rather than relying on permanent covalent bonds, Bae et al. demonstrated that data can be stored in the metastable hybridization states of secondary molecular structures. This allows the system to remain stable at room temperature but enables rapid, permanent erasure of specific blocks within minutes via mild thermal disruption.
- Epigenetic Overlays: Moving beyond canonical DNA, Wang et al. highlighted the use of non-canonical nucleic acids and epigenetic template overlays. By writing data as reversible chemical modifications (such as methylation) onto an existing nucleic acid backbone, the underlying structural template is preserved while allowing parallel data writing, modification, and erasure.
Key Foundational Work
A Rewritable, Random-Access DNA-Based Storage System (2015) — contributed 6 concept elements to this direction.
DNA-Based Storage: Trends and Methods (2015) — contributed 5 concept elements to this direction.
Portable and Error-Free DNA-Based Data Storage (2017) — contributed 3 concept elements to this direction.