Nanopore Sequencing

Nanopore sequencing is a third-generation single-molecule DNA sequencing and direct native RNA sequencing technology that determines the linear order of nucleotides—along with endogenous epigenetic base modifications—by measuring real-time picoampere fluctuations in ionic current (Δ*I*) as an unamplified single-stranded nucleic acid polymer is ratcheted electrophoretically through a nanometer-scale transmembrane protein pore or solid-state aperture. Unlike first-generation Sanger sequencing, second-generation short-read platforms, or PacBio single-molecule real-time sequencing, nanopore sequencing requires zero DNA polymerase-catalyzed synthesis, zero fluorescently labeled nucleotides, and zero optical cameras, reading the native parent DNA or RNA molecule directly.

Conceived independently in 1989 by biophysicists David W. Deamer, George M. Church, and Daniel Branton, single-stranded nucleic acid translocation across a biological pore (Staphylococcus aureus α-hemolysin) was first demonstrated experimentally by John J. Kasianowicz, Eric Brandin, Daniel Branton, and David W. Deamer in 1996 (PNAS). Following two decades of engineering advances—including high-resolution Mycobacterium smegmatis porin A (MspA) constriction discovery by Jens Gundlach and Michael Niederweis, processive enzyme motor ratcheting by Mark Akeson at UC Santa Cruz, and covalent pore chemistry by Hagan Bayley at the University of Oxford—the technology was commercialized in 2014 by Oxford Nanopore Technologies (ONT) on the pocket-sized MinION and high-throughput PromethION platforms. Because read length is governed by the physical intactness of the input DNA molecule rather than enzyme photobleaching or cycle dephasing, nanopore sequencing generates ultra-long reads exceeding 100 kilobases to >4 megabases, enabling the Telomere-to-Telomere (T2T-CHM13) Consortium to close the final centromeric and ribosomal DNA gaps of the human genome in 2022.

The Coulter-Counter Principle at the Nanoscale

A nanopore sequencing flow cell contains an array of hundreds (MinION) to thousands (PromethION) of individually addressable micro-electrodes (ASIC channels), each covered by a high-resistance synthetic amphiphilic triblock copolymer membrane separating two chambers (cis and trans) filled with potassium chloride (KCl) electrolyte buffer. Embedded within each polymer membrane is a single engineered bacterial transmembrane protein pore containing a central lumen constriction ~1.2 to 1.5 nanometers in diameter.

When a fixed direct-current bias voltage (140 to 180 mV, with the positive anode on the trans side) is applied across the insulating membrane, K⁺ and Cl⁻ ions flow steadily through the open nanopore lumen, establishing an unblocked baseline ionic current (I₀) of ~150 to 200 picoamperes (pA). Because the negatively charged phosphate backbone of single-stranded DNA or RNA carries a linear charge density of 1 e⁻ per nucleotide, the electric field captures a nucleic acid strand from the cis chamber and drives it electrophoretically through the pore toward the trans chamber. As the polymer threads through the ~1.2-nm sensing constriction, its bases sterically exclude K⁺ and Cl⁻ ions, reducing the ionic current to a characteristic residual level I* = 50 to 130 pA (sampled by an operational amplifier at 4,000 to 5,000 Hz).

Enzymatic Motor Ratcheting and Dual-Constriction Pores (R9.4.1 vs. R10.4.1)

Under an unretarded 160 mV electric field, free single-stranded DNA zips through an open α-hemolysin or MspA pore at 100,000 to 1,000,000 bases per second (<1–10 microseconds per base)—too fast for a 5-kHz electronic amplifier to resolve individual nucleotide current steps above Johnson–Nyquist thermal noise. To slow translocation by three orders of magnitude, library preparation ligates a Sequencing Adapter pre-loaded with an ATP-consuming processive motor enzyme—specifically an engineered superfamily 2 helicase (or in earlier iterations, Φ29 DNA polymerase)—to the 5′ end of the duplex DNA.

When the DNA–helicase assembly docks onto the top rim of the nanopore, the helicase unwinds the double-stranded DNA into single strands and ratchets the template strand into the pore one nucleotide at a time at a controlled velocity of ~400 to 450 bases per second (allowing ~9–12 current measurements per nucleotide step).

Even with single-base stepping, the electrical sensing zone of a biological nanopore is longer than a single 0.34-nm nucleotide:

• Single-Constriction Pore (R9.4.1): Engineered from the Escherichia coli curli transport lipoprotein CsgG, the R9.4.1 pore contains a single barrel constriction where 5 to 6 contiguous nucleotides (a 5-mer or 6-mer) simultaneously influence the ionic current (4⁵ = 1,024 to 4⁶ = 4,096 k-mer current states). When sequencing long homopolymer stretches (e.g., AAAAAAAA), a 5-mer reader head registers a flat, unchanging current once the constriction is filled with A's, making homopolymer length estimation above 5 bases error-prone.
• Dual-Constriction Pore (R10.4.1): Introduces a second engineered peptide constriction ring inside the CsgG lumen separated from the primary reader head by several nanometers. This dual-reader geometry sharpens spatial resolution in high-entropy regions while providing a secondary steric checkpoint that resolves homopolymer runs up to 9–10 bases.

From Hidden Markov Models to Recurrent and Transformer Basecallers (Dorado)

The raw output of a nanopore sequencer (stored in POD5 or FAST5 format) is a time-series trace of picoampere current steps ("squiggle" data). Because each current plateau depends on a 5- to 6-nucleotide k-mer context and motor stepping intervals follow a stochastic gamma distribution, early basecallers segmented the squiggle into discrete events and decoded transitions via a Hidden Markov Model (HMM) with Viterbi decoding (~85%–90% accuracy).

Modern nanopore bioinformatics pipelines (Guppy, Bonito, and Dorado) pass the raw unsegmented 5-kHz current trace directly into neural networks combining convolutional layers with bidirectional Long Short-Term Memory (LSTM) recurrent networks, Connectionist Temporal Classification (CTC) decoders, or Transformer self-attention layers. In Simplex mode (reading a single strand), Dorado with R10.4.1 chemistry achieves >99.2% (Q21+) modal accuracy. In Duplex mode, the complement strand is hydrodynamically tethered or follows behind the template strand through the same nanopore within seconds; joint neural decoding of the forward template squiggle and reverse complement squiggle elevates single-molecule accuracy to >99.9% (Q30+) Phred quality score.

Read-Until (Adaptive Sampling) and Direct Native RNA Sequencing

Because a nanopore reads DNA in real time at 400 bases/second and each ASIC electrode channel is independently voltage-controlled, software can basecall and align the first 400 to 800 bases (1–2 seconds of translocation) of a DNA molecule against a reference genome using GPU-accelerated minimap2 while the rest of the molecule is still threading through the pore. If the initial 500 bases map to an off-target genomic region (or host human DNA in a clinical metagenomic sample), the ASIC channel reverses its voltage polarity (from +180 mV to −180 mV) for several milliseconds, electrophoretically ejecting the unwanted DNA strand back into the cis chamber and freeing the pore to capture a new molecule. This computational target-enrichment method—Read-Until or Adaptive Sampling—enriches target genes, panels, or low-abundance pathogens by 5- to 10-fold without PCR primers or hybridization baits.

Furthermore, nanopore sequencing is capable of Direct RNA Sequencing: ligating an oligo(dT) motor adapter directly to the 3′ poly(A) tail of native messenger RNA (mRNA) or viral RNA genomes threads the intact RNA strand through the pore without reverse transcription or PCR, directly resolving full-length splice isoforms, poly(A) tail lengths, and epitranscriptomic modifications including N6-methyladenosine (m⁶A), pseudouridine (Ψ), and inosine.

Why is a processive motor enzyme (helicase) required for nanopore sequencing instead of letting the electric field pull free DNA through the pore?

Without an enzymatic motor brake, a 150-mV transmembrane potential pulls negatively charged single-stranded DNA through a 1.5-nm nanopore at ~1 to 5 microseconds per nucleotide (200,000 to 1,000,000 bases/s). At that speed, fewer than 0.02 ionic ions pass the base per step, and the 5-kHz electronic amplifier (which samples once every 200 microseconds) cannot detect individual nucleotide transitions above capacitive and thermal noise. Coupling an ATP-hydrolyzing helicase motor to the top of the pore ratchets the strand in discrete single-base steps at ~400 bases/second (~2.5 milliseconds per base), providing ~10–12 current samples per nucleotide.

What sets the maximum read length in Oxford Nanopore sequencing, and how were >4-megabase reads achieved?

Unlike Sanger, Illumina, or PacBio SMRT sequencing—where read length is bounded by electrophoretic resolution, cluster phasing, or laser-induced photodamage to a surface-immobilized polymerase—a nanopore functions as a passive sensor that reads whatever intact DNA strand enters the lumen until the physical 3′ end exits. Therefore, nanopore read length is limited only by mechanical shearing of DNA during wet-lab extraction and pipetting. By lysing cells embedded in agarose plugs or executing gentle phenol-chloroform / glass-bead precipitation (Ultra-Long DNA Sequencing Kit) with large-orifice pipette tips, researchers preserve intact chromosomal fragments spanning 100 kbp to 4.2 megabases in a single read.

How does the R10.4.1 dual-constriction nanopore improve homopolymer accuracy compared to the older R9.4.1 pore?

In the single-constriction R9.4.1 CsgG nanopore, a single ~5-nucleotide barrel constriction measures ionic current; once a homopolymer tract longer than 5 identical bases (e.g., TTTTTTTTTT) fills that single reader head, the current remains flat, forcing the neural network basecaller to estimate the tract length from the variable duration (dwell time) of the flat signal. The R10.4.1 pore incorporates a dual-constriction reader head with two spaced peptide sensing points along the translocation channel, extending the effective sensing span to ~9–10 nucleotides and providing two staggered current transitions when entering and exiting a homopolymer.

1. Kasianowicz, J. J., Brandin, E., Branton, D., & Deamer, D. W. (1996). "Characterization of Individual Polynucleotide Molecules Using a Membrane Channel." Proceedings of the National Academy of Sciences, 93(24), 13770–13773. https://doi.org/10.1073/pnas.93.24.13770
2. Cherf, G. M., Lieberman, K. R., Rashid, H., Lam, C. E., Karplus, K., & Akeson, M. (2012). "Automated Forward and Reverse Ratcheting of DNA in a Nanopore at 5-Å Precision." Nature Biotechnology, 30(4), 344–348. https://doi.org/10.1038/nbt.2147
3. Jain, M., Koren, S., Miga, K. H., Quick, J., Rand, A. C., Sasani, T. A., Tyson, J. R., Beggs, A. D., Dilthey, A. T., Fiddes, I. T., et al. (2018). "Nanopore Sequencing and Assembly of a Human Genome with Ultra-Long Reads." Nature Biotechnology, 36(4), 338–345. https://doi.org/10.1038/nbt.4060
4. Wang, Y., Zhao, Y., Bollas, A., Wang, Y., & Au, K. F. (2021). "Nanopore Sequencing Technology, Bioinformatics and Applications." Nature Biotechnology, 39(11), 1348–1365. https://doi.org/10.1038/s41587-021-01108-x