Polybarcodes in Molecular & Topological Analysis

• Polybarcodes are diverse barcode constructions that encode identities and persistent features across systems where single-barcode models are insufficient.
• In molecular assays, they attach unique DNA tags to peptides to enable high-throughput identification and precise error correction through advanced demultiplexing.
• In nanopore sequencing and applied topology, polybarcodes track structural persistence and dynamic feature reappearance, facilitating robust barcode vectorization.

Polybarcodes is a non-unified term used across several research literatures to denote barcode-based constructions that exceed the scope of a single conventional barcode. In molecular assays and sequencing, it refers to DNA tags or collections of barcodes attached to peptides, reads, or molecules, enabling highly multiplexed identification and recovery through sequencing or nanopore readout. In applied topology and symmetry theory, it denotes generalized barcode objects in which a feature may have a disconnected lifespan, so that appearance, disappearance, and reappearance are all recorded. The common theme is not a single standardized format, but the use of barcode structures to encode identity, structure, or persistence under regimes where ordinary one-bar-per-feature models are insufficient (Fink et al., 2020, Morisse et al., 2021, Liu et al., 11 Aug 2025).

1. Polybarcodes in highly multiplexed peptide assays

In peptide-based serology and related assays, polybarcodes are short, unique DNA sequences covalently attached to individual peptides in a peptide library. Each unique barcode acts as a molecular tag that uniquely identifies the corresponding peptide in the mixture. After the assay, bound peptides and thus their DNA barcodes are recovered and sequenced using high-throughput sequencing, so that the sequencing readout directly informs the identity and abundance of each peptide via its barcode. This architecture enables tracking and quantification of thousands (or more) peptides simultaneously in a single assay, and it underlies the software framework PepSIRF, which was developed specifically for the analysis of high-throughput sequencing data from highly-multiplexed peptide-based assays (Fink et al., 2020).

PepSIRF organizes polybarcode processing into modular steps. Its demux module assigns each sequencing read to a sample and to a peptide by referencing sample index sequences and a lookup of DNA tags to peptides. The search for a barcode match is performed at the expected position, then positions $\pm 1/2$ nt, and then with a configurable number of base mismatches. For libraries encoding the same peptide with multiple barcodes, counts can be consolidated at the peptide level. Downstream modules include norm for sequencing-depth adjustment, info and subjoin for subsetting and joining matrices, bin and zscore for enrichment statistics, s_enrich and p_enrich for enriched-peptide calling, and link and deconv for serological deconvolution.

The combinatorial scaling is elementary but operationally important: for a barcode of length $N$ nucleotides, the total number of unique possible barcodes is $4^N$. A plausible implication is that polybarcode assay design is governed simultaneously by library size, sequencing error, synthesis error, and the computational burden of demultiplexing. The same data record also identifies recurrent failure modes, including barcode collisions, short-barcode susceptibility to sequencing errors, and the added complexity of demultiplexing when multiple barcodes encode one peptide or when sequencing constructs are non-standard.

2. Linked-Reads, multi-barcode molecules, and robust sequencing barcode design

In Linked-Reads technologies such as 10x Genomics, barcodes tag reads which originate from a common long DNA fragment, thereby combining the high-quality and low cost of short-reads sequencing with long-range information. LRez was introduced as a C++ API and toolkit for analyzing and managing such barcode-bearing data. It supports barcode extraction from BAM files, indexing BAM and FASTQ files by barcode, querying reads or alignments sharing one or multiple barcodes, and computing the number of common barcodes between genomic regions or contig ends. Barcodes are stored using 2 bits per nucleotide, and the index maps each barcode either to file offsets or to genomic positions. The same literature notes that some libraries may use more than one barcode per molecule, and that LRez’s flexible extraction and indexing make it well-suited for such polybarcode datasets, although the paper does not elaborate specific polybarcode algorithms (Morisse et al., 2021).

These barcode operations are central to downstream genome informatics. Barcode overlap between contig ends can support scaffolding and gap filling, and barcode sharing across distant genomic regions can support structural variant calling. Query time depends on barcode abundance and can range from a couple milliseconds to a few seconds for very frequent barcodes. The tool can also run without a pre-built index for small-scale comparisons.

A separate but related line of work addresses barcode design under extreme sequencing noise. The NS-watermark method was developed for multiplex long-read sequencing with full-length single-pass reads impaired by insertion, deletion, and substitution errors in the order of $11\%$. Its construction uses an outer short LDPC code over $\mathbb{F}_q$, an inner codebook over $\mathbb{F}_4$, and a known watermark sequence added modulo $4$, yielding barcodes compatible with chemical constraints such as GC content, homopolymer length, and hairpin and dimer constraints. The method can multiplex hundreds or thousands of samples, supports barcode sets with hundreds to tens of thousands of barcodes, and achieves sample misassignment probabilities as low as $10^{-7}$; at the SMRT error profile $P_\mathrm{i}, P_\mathrm{d}=0.055$ and $P_\mathrm{s}=0.01$, the reported sample misassignment rate reaches $N$0 (Ezpeleta et al., 2016).

Taken together, these two directions distinguish between barcode management and barcode coding. One concerns extraction, indexing, and comparison in existing sequencing files; the other concerns error-resilient barcode construction before sequencing takes place.

3. Nanopore readout of barcode structure

Nanopore studies use “barcode” in a different but related physical sense: engineered protein tags, hairpins, overhangs, or other labels along a DNA molecule generate measurable current signatures as the molecule translocates through a pore. In a cylindrical nanopore, the key observable is dwell time, and in a double nanopore system the observables include dwell time, time of flight, and tag-time-delay. In both settings, Brownian dynamics simulations show that naive length estimation from local tag velocities is inaccurate because the translocating molecule does not move with a uniform velocity. The governing effect is non-equilibrium tension propagation along the chain backbone, which produces non-uniform and disparate velocities of the tags and DNA monomers (Seth et al., 2021, Seth et al., 2021).

The cylindrical nanopore study models protein tags as beads with mass three times larger and friction three times larger than standard DNA beads. It defines

$N$1

and then corrects naive tag-based distance estimates with a weighted interpolation between tag velocities and the average chain velocity. The method is recursive, uses repeated bidirectional scans, and is reported to improve barcode measurements to almost $N$2 accuracy in simulation (Seth et al., 2021).

The double nanopore study introduces a tug-of-war geometry and reversible flossing. Only the tags are practically resolvable from current blockades, and simple use of tag velocities can underestimate segment lengths, while dwell-time-based estimates may overestimate or otherwise misestimate them. The reported deviations can reach $N$3–$N$4, depending on tag spacing and location. Two corrective strategies are proposed: recalibration from a known end-to-end tag distance using a chain-average velocity, and a weighted interpolation scheme that blends tag velocities with the overall average scanning velocity. The normalized tag-time-delays also satisfy an additive sum rule,

$N$5

which provides internal consistency checks and supports inference of missing tags (Seth et al., 2021).

More recently, latent space mapping has reframed nanopore barcode analysis as a representation-learning problem rather than a time-alignment problem. A contrastive encoder trained exclusively on simulated signals from a physics-informed model maps solid-state nanopore signals of engineered DNA barcodes into an interpretable molecular coordinate system. The resulting representation is responsive to structural barcode parameters while remaining invariant to acquisition conditions and translocation conformation, enabling pooling across devices. Molecule identification requires a single pass through the encoder and reduces computational cost by three orders of magnitude relative to alignment-based methods. The reported experimental demonstrations include mixture quantification, rare-variant detection at $N$6 abundance, consensus barcode reconstruction, and real-time signal acquisition (Cartiglia et al., 15 Jun 2026).

4. Polybarcodes in persistence theory and barcode vectorization

In applied topology, polybarcodes denote generalizations of persistence barcodes that can encode richer interval structures than a standard one-interval-per-feature description. In the framework of abstract interlevel persistence for Morse-Novikov and Floer theory, the foundational objects are filtered matched pairs and Poincaré–Novikov structures. A filtered matched pair consists of a finitely generated $N$7-module together with compatible ascending and descending filtrations after extension of scalars. In the discrete case for $N$8, a doubly-orthogonal basis yields a canonical collection of bars, understood as intervals up to translation by $N$9, and the associated gaps

$4^N$0

record filtration differences for independent elements. The resulting barcodes satisfy abstract versions of stability and duality theorems; in Morse or Novikov theory they coincide with the standard barcodes coming from interlevel persistence, while in Hamiltonian Floer theory bar lengths yield quantities reminiscent of the spectral norm of a Hamiltonian diffeomorphism (Usher, 2023).

The same framework defines an essential barcode from the basis spectrum of the filtered matched pair, with intervals of the form $4^N$1 or $4^N$2 depending on the sign of $4^N$3. Stability is quantitative:

$4^N$4

Under suitable hypotheses there is also a duality that pairs closed intervals in degree $4^N$5 with open intervals in degree $4^N$6.

A related but computationally distinct development concerns vectorizing persistence barcodes for machine learning. Sara Kališnik’s work on tropical coordinates studies barcodes as unordered multisets of intervals and constructs stable max-plus and tropical rational functions on barcode space. The generators of stable max-plus polynomials depend only on interval lengths and are given by sums of the $4^N$7 longest bars, while no finite family of tropical rational functions separates all barcode equivalence classes. A countable family of tropical rational functions does separate nonequivalent barcodes, and these coordinate values can be assembled into vectors that the paper describes as polybarcodes for use in standard algorithms. These coordinates are stable with respect to both bottleneck and Wasserstein distances (Verovsek, 2016).

A common misunderstanding is that persistence polybarcodes merely rename ordinary barcodes. The literature instead uses them to handle interval data that require additional algebraic structure, translation quotients, or feature-vector embeddings beyond a plain multiset of intervals.

5. Persistent symmetries and disconnected lifespans

The most explicit recent formalization of polybarcodes appears in the study of persistent symmetries of parametrized data. For a persistence $4^N$8-configuration $4^N$9 in a metric space $11\%$0 with isometry group $11\%$1, the lifespan of a symmetry $11\%$2 is defined by

$11\%$3

The polybarcode is then

$11\%$4

Unlike a traditional barcode interval, $11\%$5 can be a finite or infinite union of intervals, possibly disjoint, so the same symmetry may appear, disappear, and later reappear. If all such lifespans are intervals, the polybarcode reduces to an ordinary barcode (Liu et al., 11 Aug 2025).

This framework is designed to capture birth, death, persistence, and reappearance of symmetries over parameter evolution. It defines distances on closed subsets of $11\%$6, including symmetric-difference and expansion-style metrics, and proves a stability theorem of the form

$11\%$7

The same work places polybarcodes within a larger program that includes persistent symmetry groups, symmetry defect, persistence representations of persistence groups, and persistent Fourier analysis on persistence groups, with intended applications to symmetry breaking and phase transitions.

Here the distinction from classical persistent homology is exact rather than metaphorical. Traditional barcodes encode connected intervals for homological features; symmetry polybarcodes encode possibly non-connected supports for individual symmetries. The ability to record reappearance is the central conceptual difference.

6. Adjacent barcode literatures and terminological boundaries

Several adjacent barcode literatures are sometimes conflated with polybarcodes but study different objects. DNA2DBC is a two-dimensional barcode symbology for encoding DNA and text metadata into a square matrix. It uses 3-bit codewords for DNA symbols, in-band mode switching between DNA and text modes, a square layout with solid and dotted borders for alignment, and Reed-Solomon error correction adapted to modulo-$11\%$8 codewords. The row and column count is determined by

$11\%$9

Its goal is machine-readable physical storage of DNA information rather than multiplex molecular identification or persistent feature tracking (Purcaru et al., 2012).

Polymer-based data storage provides another neighboring but distinct usage domain. There, the problem is not a visible or sequencing barcode, but coding over chains of binary synthetic polymers read via tandem mass spectrometry. The central tasks are unique string reconstruction from erroneous substring composition multisets and correction of multiple mass errors. The constructions use Catalan-Bertrand paths, Reed-Solomon-style redundancy in the asymmetric case, and polynomial evaluation codes in the symmetric case. The asymptotic code rate is one, with redundancy scaling linearly in the number of asymmetric errors and quadratically in the number of symmetric errors, in both cases logarithmically in string length (Pattabiraman et al., 2020).

This broader context suggests that “polybarcodes” functions less as a single standardized technical noun than as a family resemblance across barcode-intensive problems. In some papers it denotes multiple barcodes per molecule; in others it denotes generalized persistence objects or vectors of barcode coordinates; and in yet others it is adjacent to, but not synonymous with, robust coding and decoding frameworks for DNA-like media. The literature therefore supports a plural definition: polybarcodes are barcode constructions designed for settings in which multiplicity, reappearance, high error, or structural ambiguity make ordinary barcode models insufficient.