Third generation sequencing is a class of DNA-reading technology that analyzes single molecules of DNA or RNA in real time, without the need to chop them into tiny fragments and amplify them first. The two dominant platforms, Pacific Biosciences (PacBio) and Oxford Nanopore Technologies (ONT), each use a fundamentally different detection method, but both share the defining trait: they can read long, continuous stretches of a single molecule, routinely producing reads tens of thousands of bases long and sometimes exceeding a million. This matters because earlier sequencing methods relied on reading millions of short snippets and computationally stitching them back together, a process that struggles with repetitive regions and complex structural changes in genomes. Third generation sequencing has already enabled the first truly complete sequence of a human genome and is reshaping clinical diagnostics, but the technology comes with trade-offs worth understanding.
How It Differs From Earlier Sequencing
First generation sequencing, the Sanger method developed in the 1970s, reads one fragment at a time and was the workhorse behind the original Human Genome Project. Second generation sequencing, often called next generation sequencing or NGS, dramatically scaled up throughput by running millions of reactions in parallel on a chip. Platforms like Illumina dominate this space and remain the backbone of most genomics labs. NGS is fast, cheap per base, and extremely accurate on a per-read basis, but its reads are short, typically under 300 base pairs. That shortness creates blind spots in repetitive stretches, centromeres, and regions packed with duplicated segments.
Third generation sequencing was built to solve those blind spots. Rather than amplifying DNA into clusters and reading synthetic copies, it watches a single native molecule being processed. Because no amplification step is needed, the original chemical modifications on the DNA (like methylation) are preserved, which opens the door to reading genetic and epigenetic information simultaneously. The trade-off historically has been higher error rates and greater cost per base compared to short-read platforms.1PubMed Central. DBG2OLC: Efficient Assembly of Large Genomes Using Long Erroneous Reads of the Third Generation Sequencing Technologies
PacBio SMRT Sequencing
PacBio’s method is called Single Molecule, Real-Time (SMRT) sequencing. The core hardware is a chip containing thousands of tiny wells called zero-mode waveguides, or ZMWs. Each well is far smaller than the wavelength of visible light, which creates an unusual optical property: only light emitted right at the bottom of the well can be detected. A single DNA polymerase enzyme is fixed to the bottom of each ZMW, and it processes one molecule of DNA at a time.2PubMed Central. Zero-mode waveguides and nanopore-based sequencing technologies accelerate single-molecule studies
As the polymerase incorporates nucleotides (the building blocks of DNA), each of the four types carries a different fluorescent label. A camera beneath the chip records the color and timing of each flash of light. The sequence of flashes reveals the sequence of bases. Because the polymerase works on a circular DNA template, it can loop around the same molecule multiple times, generating what are called “subreads.” Averaging those subreads together produces a consensus read, called a circular consensus sequence or CCS read, with much higher accuracy than any single pass. PacBio recommends library insert sizes of roughly 10 to 20 kilobases for this mode, which means each consensus read covers a substantial stretch of genome in one shot.3Nature Communications. High-throughput and high-accuracy single-cell RNA isoform analysis using PacBio circular consensus sequencing
PacBio’s high-fidelity (HiFi) reads, the product of this consensus approach, now achieve accuracy above 99.9 percent, putting them in the same ballpark as short-read platforms while retaining the advantage of long reads. This combination of length and accuracy has made HiFi reads a go-to tool for assembling complex genomes.
Oxford Nanopore Sequencing
Oxford Nanopore Technologies takes a completely different approach. Instead of watching a polymerase copy DNA, nanopore sequencing threads a single strand of DNA or RNA directly through a tiny protein pore embedded in a membrane. An electric current flows through the pore, and as each base passes through, it partially blocks the current in a characteristic way. By measuring those fluctuations in real time, the system reads the sequence of bases.
Because the method is reading native molecules, there is no theoretical upper limit on read length. Researchers have produced individual reads exceeding two million bases. The platform also sequences RNA directly, without converting it to DNA first, which avoids the biases introduced by reverse transcription.4PubMed Central. An Introduction to Nanopore Sequencing: Past, Present, and Future Considerations ONT devices range from the pocket-sized MinION, roughly the size of a large USB stick, to the high-throughput PromethION for production-scale work. This scalability and portability have made nanopore sequencing popular in settings far from traditional labs.
The Accuracy Question
Early third generation sequencing had a reputation for sloppiness. Raw single-pass error rates above 10 percent were common, which made the technology unsuitable for applications where every base call matters. That picture has changed dramatically in the past few years on both platforms.
For PacBio, the circular consensus approach largely solved the accuracy problem. By reading the same molecule multiple times and building a consensus, HiFi reads achieve single-molecule accuracies comparable to short-read data. For nanopore sequencing, accuracy improvements have come from better pore chemistry, faster electronics, and more sophisticated computational models (basecallers) that translate raw electrical signals into base sequences. ONT’s R10.4 flowcell and updated chemistry produce per-read accuracies that rival Illumina data, especially when using their highest-accuracy basecalling models and duplex reads, where both strands of a double-stranded molecule are sequenced.5PubMed Central. Comparison of R9.4.1/Kit10 and R10/Kit12 Oxford Nanopore flowcells and chemistries in bacterial genome reconstruction
More recent upgrades have pushed nanopore error rates even lower. With V14 chemistry, total error rates below 0.05 percent have been demonstrated for amplicon sequencing, and updated basecallers trained on native bacterial DNA have reduced systematic errors like specific base substitutions that previously plagued methylated regions.6PubMed Central. Oxford Nanopore enhanced accuracy of long-read amplicons applied to microbial whole-genome sequencing 7PubMed Central. Evaluation of the accuracy of bacterial genome reconstruction with Oxford Nanopore R10.4.1 long-read-only sequencing The gap between long-read and short-read accuracy, once a dealbreaker, has narrowed to the point where many labs now use nanopore or PacBio sequencing alone for tasks that previously required short-read data as a quality check.
Why Long Reads Matter for Structural Variants
One of the clearest advantages of third generation sequencing is its ability to detect structural variants, the large-scale rearrangements in a genome that include deletions, duplications, inversions, and insertions of hundreds or thousands of bases. Short-read sequencing struggles with these because the reads are shorter than many of the variants themselves. Imagine trying to detect a paragraph that has been moved to a different chapter in a book when you can only see one sentence at a time. Long reads span entire variants and their surrounding context, making detection far more reliable.
This capability is transforming the study of rare genetic disorders, where structural variants account for a meaningful share of disease-causing mutations that go undetected by standard short-read methods. Long-read platforms can now accurately find structural variants even in previously unreachable parts of the genome, including repetitive sequences and segmental duplications that short reads simply cannot resolve.8PubMed Central. Long-Read Sequencing and Structural Variant Detection: Unlocking the Hidden Genome in Rare Genetic Disorders
Completing the Human Genome
The original Human Genome Project, declared “complete” in 2003, actually left about 8 percent of the genome unsequenced. Those gaps lived in the most repetitive, hardest-to-assemble regions: centromeres, the short arms of certain chromosomes, and areas packed with near-identical segmental duplications. Short reads could not bridge them.
In 2022, the Telomere-to-Telomere (T2T) Consortium published the first truly gapless sequence of a human genome, a 3.055-billion-base-pair assembly called T2T-CHM13. That effort added nearly 200 million base pairs of previously missing sequence, including 1,956 gene predictions and 99 predicted protein-coding genes that had never been catalogued.9PubMed Central. The complete sequence of a human genome The project relied heavily on ultra-long Oxford Nanopore reads to span the most complex repeats and PacBio HiFi reads to provide a high-resolution assembly graph. Tools like Verkko were later developed to automate and improve this hybrid strategy for assembling complete diploid genomes.10PubMed Central. Telomere-to-telomere assembly of diploid chromosomes with Verkko
Completing the human genome was not just symbolic. Those newly resolved regions include all centromeric satellite arrays and the short arms of all five acrocentric chromosomes, areas now open to variational and functional studies for the first time. The achievement would not have been possible without long-read sequencing.
Reading Epigenetics Without Extra Steps
In traditional sequencing workflows, detecting chemical modifications on DNA, such as methylation, requires a separate treatment step (like bisulfite conversion) that chemically alters the DNA before sequencing. This adds time, cost, and can introduce artifacts. Third generation sequencing can detect these modifications natively, as a natural byproduct of reading the molecule.
PacBio’s SMRT sequencing detects methylation by measuring how the polymerase’s speed changes when it encounters a modified base. The arrival times and durations of fluorescence pulses shift in characteristic ways for different modifications, allowing the system to identify N6-methyladenine, 5-methylcytosine, and 5-hydroxymethylcytosine without any chemical pretreatment.11PubMed Central. Direct detection of DNA methylation during single-molecule, real-time sequencing Nanopore sequencing achieves something similar: modified bases alter the current signal as they pass through the pore, and trained computational models can distinguish them from unmodified bases.
This ability to read the genome and its epigenome in one pass is genuinely new. Methylation patterns play central roles in gene regulation, development, and disease, and being able to map them alongside the DNA sequence at the single-molecule level has opened research questions that were previously impractical to tackle at scale.
Direct RNA Sequencing
Most RNA sequencing involves converting RNA to complementary DNA (cDNA) through reverse transcription, then sequencing the cDNA. This conversion step introduces biases: some transcripts are copied more efficiently than others, information about RNA modifications is lost, and the process can create chimeric artifacts. Oxford Nanopore’s direct RNA sequencing method skips this entirely, threading native RNA molecules through the pore. The result is full-length, strand-specific RNA sequences with the ability to detect nucleotide analogs directly.12PubMed Central. Highly parallel direct RNA sequencing on an array of nanopores
For researchers studying alternative splicing, RNA modifications, or transcript diversity, direct RNA sequencing removes an entire layer of potential distortion. It is still lower-throughput than cDNA-based approaches, but it provides a kind of ground truth about what RNA molecules actually look like in a cell.
Sequencing in the Field
The MinION’s portability has enabled sequencing in places that would have been unthinkable a decade ago. Researchers have used it during Ebola and Zika outbreaks to perform rapid pathogen surveillance in remote clinics, aboard the International Space Station, and in field stations monitoring environmental DNA. One group developed a portable pipeline using the MinION to detect harmful algal blooms in Lake Erie by sequencing environmental DNA on site, producing rapid, low-cost results that could inform public health decisions in near-real time.13bioRxiv. Environmental DNA sequencing data from algal blooms in Lake Erie using Oxford Nanopore MinION
The device runs off a laptop’s USB port, needs no special infrastructure, and can start returning sequence data within minutes of loading a sample. For applications where speed and location flexibility matter more than raw throughput, nanopore sequencing has created a category that did not previously exist.
Clinical Diagnostics and Speed
Speed is critical in clinical genomics, especially for critically ill patients where a diagnosis can change treatment within hours. Long-read sequencing is proving its worth here. In one study of critically ill patients undergoing nanopore-based long-read genome sequencing, a genetic diagnosis was achieved for about 42 percent of participants, with an average turnaround of 5.3 days from sample receipt to result, compared to 18.4 days under standard genomic care. Of those who received a genetic diagnosis, roughly two-thirds had significant and immediate changes to their clinical management.14European Journal of Human Genetics. Nanopore long-read sequencing for the critically ill facilitates ultrarapid diagnostics and urgent clinical decision making
A separate study of critically ill children in Thailand used rapid long-read genome sequencing as a first-tier test and achieved a diagnostic rate of 61 percent with a median turnaround of nine days. The long reads were particularly valuable for identifying large structural deletions that short-read sequencing would have missed, and for phasing variants, determining which copy of a gene each variant sits on, which helped reclassify ambiguous variants as disease-causing.15PubMed Central. Singleton rapid long-read genome sequencing as first tier genetic test for critically Ill children with suspected genetic diseases All eleven patients who received a diagnosis had changes to their clinical management as a result. These early studies suggest that long-read sequencing can serve as a single comprehensive test, replacing the multiple sequential tests that standard workflows often require.
Hybrid Approaches
Long-read and short-read technologies are not mutually exclusive. Many projects combine both in what is called hybrid assembly. Short reads from Illumina provide highly accurate base-level information, while long reads from PacBio or ONT span repetitive regions and resolve the genome’s overall architecture. The combination produces high-quality assemblies at reduced cost compared to using either technology alone.16Bioinformatics. hybridSPAdes: an algorithm for hybrid assembly of short and long reads This approach has been especially productive for bacterial genomes, where even modest amounts of long-read data combined with short reads can resolve plasmids and repetitive elements that short reads alone cannot untangle.17PubMed Central. Comparison of long-read sequencing technologies in the hybrid assembly of complex bacterial genomes
As long-read accuracy has improved, the need for short-read polishing has diminished for some applications. But hybrid strategies remain popular when budgets are tight or when the highest possible accuracy is non-negotiable, such as in clinical variant calling.
Single-Cell and Spatial Applications
One of the more exciting frontiers is pairing long-read sequencing with single-cell and spatial transcriptomics. Standard single-cell RNA sequencing captures which genes are active in individual cells but typically uses short reads that cannot distinguish between different isoforms, the variant forms of a gene’s RNA that result from alternative splicing. Long-read methods can capture full-length transcripts, revealing not just which genes are on but which specific versions of those genes each cell is producing.18PubMed Central. Understanding isoform expression by pairing long-read sequencing with single-cell and spatial transcriptomics
Spatial long-read approaches take this further by preserving the physical location of transcripts within a tissue. One recent method, called Spl-ISO-Seq, achieves near-single-cell resolution while doubling to tripling read lengths compared to standard preparations, making it possible to study how splicing and polyadenylation patterns vary across different layers of the brain cortex, for example.19Nature Communications. A spatial long-read approach at near-single-cell resolution reveals developmental regulation of splicing and polyadenylation sites in distinct cortical layers and cell types
Pangenomics and Population-Scale Diversity
For decades, genomics has relied on a single linear reference genome, essentially a composite from a handful of individuals, to which new data is compared. This creates a built-in bias: genetic variants that do not exist in the reference are harder to detect, and populations underrepresented in the reference are systematically disadvantaged. Long-read sequencing has been central to the shift toward pangenome references, graph-based structures that represent the full spectrum of human variation rather than a single path through the genome.
Projects like the Human Pangenome Reference Consortium have used long-read data to identify hundreds of megabases of genetic diversity that were missing from the old linear reference, leading to substantial improvements in how structural variants are detected and how haplotypes are reconstructed across diverse populations.20PubMed Central. Beyond single references: pangenome graphs and the future of genomic medicine This work is still in progress, but the trajectory is clear: long reads are making genomics more inclusive and more complete.
Ancient DNA and Nanopore Sequencing
An unexpected application of nanopore sequencing is in the study of ancient DNA. Archaeological and paleontological samples contain DNA that is heavily degraded, fragmented, and chemically damaged. These characteristics make ancient DNA a poor fit for platforms that require long, intact molecules. But nanopore sequencing’s ability to read very short fragments, combined with its portability, has attracted interest from the paleogenomics community.
Researchers have tested nanopore sequencing on ancient samples and confirmed that the reads display the hallmark damage profiles expected of authentic ancient DNA, including deamination-derived substitutions accumulating toward the ends of reads. Average read lengths from such samples are short (around 79.5 base pairs in one study), reflecting the degraded state of the input material, but the technology can still recover usable sequence data. The appeal is the possibility of performing initial screening or even preliminary analysis on site at an excavation, rather than shipping samples to a centralized lab.21bioRxiv. Towards On-Site Paleogenomics: Application and Perspective of Nanopore Sequencing with Ancient DNA
Computational Demands
Long reads come with computational challenges that are easy to underestimate. Basecalling, the step where raw electrical or optical signals are converted into DNA sequences, is computationally expensive, especially for nanopore data where the highest-accuracy models require significant GPU processing power. Mapping long, error-prone reads to a reference genome also demands different algorithms than those optimized for short reads, and existing methods can be slow on large genomes.22Bioinformatics. Real-time mapping of nanopore raw signals
For small labs or field deployments, the computational bottleneck can be as limiting as the sequencing itself. Cloud computing has helped, but real-time analysis of nanopore data in remote settings remains a work in progress. The bioinformatics ecosystem around long reads is maturing rapidly, with new tools appearing frequently, but users switching from short-read workflows should expect a learning curve and higher compute requirements.
Comparing PacBio and Nanopore
The two platforms occupy overlapping but distinct niches. Head-to-head comparisons generally find that PacBio produces higher data quality, while ONT provides higher yield and greater flexibility in read length. For transcriptome analysis, PacBio has performed marginally better in most metrics for both long-read-only and hybrid strategies.23PubMed Central. Comprehensive comparison of Pacific Biosciences and Oxford Nanopore Technologies and their applications to transcriptome analysis That said, ONT’s portability, lower upfront hardware cost, and ability to generate ultra-long reads give it advantages in field work, rapid clinical turnaround, and applications that need reads spanning very large structural features. Many large-scale projects use both: PacBio HiFi for accuracy-critical assembly and ONT ultra-long reads for spanning the trickiest repetitive regions, exactly as the T2T Consortium did.
The choice between platforms often comes down to the specific question being asked, the available budget, and whether portability matters. Neither platform is universally superior, and the competitive pressure between them has driven remarkably fast improvements on both sides.
Solid-State Nanopores and What Comes Next
Current nanopore sequencing relies on biological protein pores, which have limitations in durability and the range of conditions they can tolerate. Solid-state nanopores, fabricated from synthetic materials like silicon nitride or graphene, offer the promise of greater stability, tunability, and integration with semiconductor manufacturing processes. Researchers have been investigating solid-state nanopore sequencing intensively, though significant challenges remain in achieving the single-base resolution needed for practical sequencing.24PubMed Central. Computational studies of DNA sequencing with solid-state nanopores: key issues and future prospects
Other emerging directions include electronic detection methods that could bypass optical systems entirely, further miniaturization of devices, and tighter integration of sequencing with on-chip sample preparation. Whether these constitute a “fourth generation” or simply the continued evolution of third generation technology is mostly a question of marketing. What matters is the trajectory: sequencing is getting longer, more accurate, cheaper, more portable, and more capable of reading biological information beyond just the sequence of bases.