Oxford Nanopore Technology (ONT) sequencing reads DNA or RNA by threading single strands through a tiny protein pore and measuring the electrical disruptions each stretch of sequence causes as it passes through. Unlike older sequencing methods that work by copying DNA and reading the copies, ONT reads the original molecule directly, in real time, and can handle fragments hundreds of thousands of bases long. That combination of directness, speed, and read length has made it one of the most versatile tools in modern genomics, but how it actually pulls this off is more interesting than most summaries let on.
The Nanopore Itself
At the heart of the technology is a nanopore, a protein channel embedded in an electrically resistant membrane. When a voltage is applied across the membrane, ions flow through the pore, creating a steady electrical current. A single strand of DNA (or RNA) is then fed through this channel, and as each short stretch of nucleotides sits inside the pore’s narrowest point, it partially blocks the ion flow. Different combinations of bases block the current by different amounts, producing a characteristic signal for each sequence context.
The pore used in most ONT devices is a modified version of a bacterial protein. Early research characterized the system using a porin from Mycobacterium smegmatis called MspA, paired with a helicase enzyme (Hel308) that acts as a molecular ratchet. The helicase grabs hold of the DNA strand and steps it through the pore one nucleotide at a time, controlling the speed so the electrical signal can be recorded at each position. Each step actually involves two distinct sub-movements corresponding to open and closed conformations of the helicase as it binds and breaks down ATP for energy.1Nature Communications. An information theory approach to quantifying the sequence-dependent response of nucleic acid motors with applications to nanopore DNA sequencing Without this motor protein, the DNA would zip through the pore too fast to read.
The kinetics of this stepping process are not uniform. The helicase’s speed and even its tendency to occasionally slip backwards depend on the specific DNA sequence passing through it. Research has traced this sequence dependence to about ten sites where the helicase makes direct contact with individual DNA bases, with two positions in particular exerting the strongest influence on translocation behavior.2Nucleic Acids Research. Determining the effects of DNA sequence on Hel308 helicase translocation along single-stranded DNA using nanopore tweezers This means the raw signal coming out of the pore reflects not just what bases are present, but how the motor protein interacts with those bases, adding a layer of complexity to the data.
From Electrical Signal to DNA Sequence
The raw output of a nanopore run is not a neat string of A’s, T’s, G’s, and C’s. It is a noisy, squiggly trace of current measurements over time. Translating that trace into an actual sequence is the job of software called a basecaller, and it is one of the most computationally demanding parts of the whole process.
Modern basecallers rely on neural networks trained on real sequencing data. The network learns to associate patterns in the electrical signal with specific sequences. Early versions used a two-step process: first segment the signal into discrete “events” (one per nucleotide step), then call bases from those events. A major leap came when basecallers switched to working directly on the raw, unsegmented signal, which improved accuracy and especially helped with sequences where the same base repeats multiple times in a row.3PubMed Central. Performance of neural network basecalling tools for Oxford Nanopore sequencing Bigger, more complex neural networks deliver better accuracy, though they take longer to run.
The architecture of these networks has grown more sophisticated over time. Some designs stack convolutional layers (which pick up local patterns in the signal) with recurrent layers (which capture dependencies across longer stretches), then feed the combined output through a decoding framework that handles the mismatch between signal length and sequence length.4Bioinformatics. MSRCall: a multi-scale deep neural network to basecall Oxford Nanopore sequences The practical upshot for users is that the same raw data can be re-basecalled later with newer, better software and yield a more accurate sequence without running the experiment again.
What the signal encodes is itself revealing. Analysis of how individual nucleotides influence the current has shown that the identity of the base currently sitting in the pore’s constriction matters most, but nearby bases also shift the signal. For instance, a thymine at certain positions within the reading window consistently pushes the current higher, while other bases and positions have subtler or opposite effects.5PubMed Central. Understanding the Impact of Individual Nucleotide on Oxford Nanopore Current Signals With Interpretable Prediction Models The current at any given moment reflects a “k-mer,” a short window of roughly five to nine bases occupying the pore simultaneously, rather than a single isolated nucleotide.
How Accurate Is It?
Accuracy has been the persistent question mark around nanopore sequencing since its earliest days, and the honest answer is that it has improved dramatically but still trails the best short-read platforms on a per-read basis. The trajectory, though, has been steep. ONT’s newer R10.4 flow cells produce reads with an average accuracy of roughly 97% and a modal accuracy around 99%, a clear jump over the older R9.4.1 chemistry, which sat a couple of percentage points lower.6Computational and Structural Biotechnology Journal. Benchmarking of Nanopore R10.4 and R9.4.1 flow cells in single-cell whole-genome amplification and whole-genome shotgun sequencing Much of that improvement comes from better handling of homopolymers, stretches where the same base repeats several times in a row.
At the assembly level, R10.4 data has proven good enough to produce near-finished bacterial genomes without any help from a second sequencing technology. Earlier versions essentially required “polishing” with high-accuracy short reads to clean up systematic errors, but R10.4 assemblies showed no significant improvement when short-read polishing was added, suggesting the remaining errors are sparse and random rather than systematic.7Nature Methods. Oxford Nanopore R10.4 long-read sequencing enables the generation of near-finished bacterial genomes from pure cultures and metagenomes without short-read or reference polishing For a technology that was considered rough around the edges just a few years ago, that is a meaningful milestone.
Where Errors Still Cluster
The errors that remain in nanopore data are not evenly distributed. They concentrate in specific sequence contexts, and understanding those patterns matters for anyone interpreting results.
The most well-known trouble spot is homopolymers. When several identical bases line up in a row, the signal barely changes from one step to the next, making it hard for the basecaller to determine exactly how many bases are present. The usual mistake is underestimating the length, producing deletion errors. Roughly half of all sequencing errors in nanopore DNA data trace back to homopolymers.8PLoS ONE. Sequencing DNA with nanopores: Troubles and biases Software tools specifically designed to correct these systematic errors using comparisons to related reference sequences have shown substantial improvements over general-purpose correction methods.9PubMed Central. Homopolish: a method for the removal of systematic errors in nanopore sequencing by homologous polishing
Direct RNA sequencing introduces its own error landscape. In native RNA data, errors are more evenly spread between homopolymeric and non-homopolymeric regions. Short homopolymers (two or three bases) and mixed-sequence stretches together account for over 90% of the sequenced nucleotides and show similar deletion and mismatch rates. Longer homopolymers are more error-prone per base but are rare enough that they do not dominate the overall error profile the way they do in DNA data.10PubMed Central. Sequencing accuracy and systematic errors of nanopore direct RNA sequencing
Reading RNA and Detecting Chemical Modifications
One of the genuinely distinctive capabilities of nanopore sequencing is that it can read RNA molecules directly, without first converting them to DNA copies. Traditional RNA sequencing requires a reverse transcription step that introduces biases, loses information about chemical modifications on the original RNA, and often fragments the molecule into short pieces. Nanopore direct RNA sequencing sidesteps all of that: a full-length RNA strand passes through the pore, and the resulting signal captures both the sequence and any chemical modifications present on the bases.11Nature Methods. Nanopore native RNA sequencing of a human poly(A) transcriptome The method yields strand-specific, full-length reads and can detect nucleotide analogs in real time.12PubMed Central. Highly parallel direct RNA sequencing on an array of nanopores
The same principle applies to DNA methylation, one of the most studied chemical modifications in the genome. Because ONT reads native DNA rather than amplified copies, methyl groups on cytosines produce recognizable distortions in the electrical signal. Current basecalling software can call both 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC) simultaneously, which matters because the two modifications have different biological roles and can be confused with each other if only one is modeled.13Scientific Reports. Reliable investigation of DNA methylation using Oxford nanopore technologies Getting sequence and methylation data from the same molecule in the same run is something no copy-based method can do.
Ultra-Long Reads and What They Unlock
If accuracy has been the technology’s Achilles heel, read length is its superpower. While short-read sequencers produce fragments of a few hundred bases, ONT routinely generates reads tens of thousands of bases long, and optimized protocols have pushed individual reads past five million bases. One study achieved average N50 read lengths of about 80 kilobases and maximum individual reads of 5.83 megabases, with extracted DNA fragments exceeding 485 kilobases.14PubMed. Nanopore ultra-long sequencing and adaptive sampling spur plant complete telomere-to-telomere genome assembly
These ultra-long reads are not just a party trick. Genomes are full of repetitive sequences, sometimes spanning tens or hundreds of kilobases, that short reads cannot bridge. When a read is longer than the repeat, the assembler can place it unambiguously. This capability has been central to the push toward telomere-to-telomere genome assemblies, where every base from one end of a chromosome to the other is resolved. Long nanopore reads are also better at detecting structural variants, the large-scale rearrangements, insertions, and deletions that short reads often miss entirely. Research has shown that nanopore long reads outperform short reads for finding complex rearrangements and can phase variants to determine which parent they came from.15Nature Communications. Mapping and phasing of structural variation in patient genomes using nanopore sequencing Many of the novel variants discovered by long reads turn out to be retrotransposon insertions invisible to standard short-read pipelines.
For rare genetic disorders, the ability to see structural variants in previously inaccessible parts of the genome, including repetitive sequences and segmental duplications, is opening up diagnoses that were not possible before.16PubMed Central. Long-Read Sequencing and Structural Variant Detection: Unlocking the Hidden Genome in Rare Genetic Disorders
Adaptive Sampling
A feature unique to nanopore sequencing is the ability to make accept-or-reject decisions on individual molecules while the run is still happening. During adaptive sampling, the sequencer reads the first few hundred bases of each strand, compares them in real time to a region of interest, and either keeps reading or reverses the voltage to eject the strand from the pore. Ejected strands make room for new molecules, effectively enriching the data for your target region without any upfront laboratory enrichment step.
The enrichment is meaningful. In one evaluation, channels performing adaptive sampling achieved roughly 17-fold coverage of the target region compared to about 5-fold from standard sequencing, a roughly 3.6-fold improvement from software alone.17bioRxiv. Tech Note – Adaptive Sampling: targeted Oxford Nanopore long-read sequencing This has practical implications for clinical settings where you might want to deeply sequence a specific gene panel or a suspicious region without needing to design and purchase capture probes.
Portability and Field Sequencing
The smallest ONT device, the MinION, is roughly the size of a stapler and plugs into a laptop via USB. This form factor has made genomic sequencing possible in places no one would have imagined a sequencing lab could go.
The most famous demonstration was during the 2014-2016 Ebola outbreak in West Africa. Researchers packed a complete nanopore sequencing system into standard airline luggage, flew it to Guinea, and set up real-time genomic surveillance of the virus. They generated sequence data from Ebola-positive samples in under 24 hours, with the sequencing step itself taking as little as 15 to 60 minutes.18PubMed Central. Real-time, portable genome sequencing for Ebola surveillance That kind of turnaround in a resource-limited setting was unprecedented.
Field use has since been pushed to more extreme environments. A team successfully ran nanopore sequencing in the Antarctic dry valleys, one of the harshest environments on Earth, by insulating the sequencer and laptop well enough to sustain runs for up to two and a half hours under arduous conditions.19PubMed Central. Real-Time DNA Sequencing in the Antarctic Dry Valleys Using the Oxford Nanopore Sequencer Other groups have tested the technology for field-based biothreat detection, transporting flow cells and reagents on commercial flights despite imperfect temperature control and pressure changes.20PLOS ONE. Field-based detection of bacteria using nanopore sequencing: Method evaluation for biothreat detection in complex samples The devices are rugged enough to tolerate conditions that would wreck a traditional sequencer, though performance can degrade when reagents are handled roughly.
Microbial Identification and Metagenomics
One area where nanopore’s long reads pay off in a very practical way is identifying microbes. The standard genetic barcode for bacteria, the 16S ribosomal RNA gene, is about 1,500 bases long. Short-read sequencers can only capture fragments of it, which limits how precisely you can identify a species. Nanopore sequencing can read the entire gene in one pass, providing more precise taxonomic classification.21PubMed Central. Microbial Community Profiling Protocol with Full-length 16S rRNA Sequences and Emu
A comparative analysis across three sequencing platforms found that ONT achieved the best species-level resolution for gut microbiota, classifying 76% of sequences to species, compared to 63% for PacBio and 48% for Illumina short reads. The caveat is that across all platforms, a large fraction of species-level labels were “uncultured bacterium,” reflecting gaps in reference databases rather than sequencing technology limitations.22PubMed Central. Comparative analysis of Illumina, PacBio, and nanopore for 16S rRNA gene sequencing of rabbit’s gut microbiota The sequencer can read the full gene beautifully, but if nobody has cultured and characterized the organism before, there is no name to match it to.
Protein Sequencing on the Horizon
The nanopore concept is not limited to nucleic acids. Researchers have been exploring whether the same basic principle, threading a linear polymer through a pore and reading the electrical signal, could work for proteins. The challenges are steeper: proteins fold into complex three-dimensional shapes that would not fit through a narrow pore, and unlike DNA, proteins do not carry a uniform electrical charge that would drive them through in one direction.
Solid-state nanopores, synthetic versions made from materials like silicon nitride rather than biological proteins, have shown promise for detecting and discriminating individual protein molecules based on their size and shape.23PubMed Central. Application of Solid-State Nanopore in Protein Detection To address the folding and charge problems, one approach uses the detergent SDS (the same compound found in many shampoos) to unfold proteins and coat them with negative charge. Molecular simulations and experiments have confirmed that SDS-treated proteins lose their folded structure during translocation and move through the pore in the direction dictated by the electric field, both prerequisites for eventual protein sequencing.24Nanoscale. SDS-assisted protein transport through solid-state nanopores True single-molecule protein sequencing by nanopore remains a research goal rather than a product, but the groundwork is being laid.