For over a decade, next-generation sequencing (NGS) has dominated genomics by reading DNA in short, highly accurate fragments of 100 to 300 base pairs. While highly effective for identifying single-nucleotide mutations, short-read sequencing suffers from a fundamental blind spot: structural variants (SVs). SVs—which include large deletions, duplications, inversions, translocations, and mobile element insertions—frequently span thousands to millions of base pairs.
Trying to reconstruct a large structural rearrangement from short reads is like trying to assemble a massive jigsaw puzzle of a clear blue sky. If a mutation occurs within a highly repetitive genomic region or involves a massive shuffling of chromosomal pieces, short fragments cannot be uniquely mapped back to a reference genome. The critical context—the "breakpoints" where the DNA sequence deviates from the norm—is lost because the individual reads are simply too short to span the variation and anchor themselves into unique, known genomic territory.
Long-read sequencing (LRS), often referred to as third-generation sequencing (TGS), fundamentally changes this paradigm. By reading continuous strands of DNA that span tens of thousands to millions of base pairs, these technologies capture the entire structural variant and its flanking regions in a single, unbroken molecule. This eliminates the ambiguity of fragment reassembly, allowing researchers to directly observe the architecture of complex genomes.
Short-Read Sequencing (NGS)
DNA Fragment: [---] [---] [---]
Breakpoint: X (Missed or unmappable due to repetitive context)
Long-Read Sequencing (TGS)
Continuous Read: [==================X==================]
Breakpoint: ^ Captured cleanly with unique flanking context
Unmasking the Non-Coding Genome and Chromatin Topology
A significant portion of human genetic diversity and disease burden lies outside of protein-coding exons. In these non-coding regions, structural variations often disrupt regulatory elements like enhancers or alter the three-dimensional folding of the genome (chromatin topology). Because non-coding and regulatory regions are frequently rich in repetitive elements, short-read methods routinely miss these pathogenic variants.
Long-read sequencing overcomes these limitations by providing the continuity needed to map repetitive non-coding spaces. For instance, Wenmin Sun et al. (2020) utilized long reads to accurately characterize large deletions in a non-coding region of chromosome 1q32.1, identifying them as the root cause of Congenital Motor Nystagmus in two large families. Without long reads to span the breakpoints, these deletions remained hidden within the complex architecture of the non-coding genome.
Beyond simple deletions, long reads are crucial for resolving balanced translocations and inversions that shuffle the genome without changing the total amount of DNA. In a study on central iris hypoplasia, Wenmin Sun et al. (2024) demonstrated how a balanced chromosomal translocation disrupted the 3D genome structure. By using long-read sequencing to pinpoint the exact structural boundaries, the authors showed that the rearrangement altered chromatin topology, forcing aberrant interactions between distant regulatory elements and the APCDD1 gene.
This capacity to span complex regions also illuminates human evolution. By applying single-molecule long-read sequencing to Tibetan and Han individuals, Jinlong Shi et al. (2023) uncovered a previously invisible landscape of structural variation. They identified a specific deletion that disrupts a super-enhancer for the EPAS1 gene, providing a definitive mechanical explanation for how Tibetans adapted to low-oxygen environments—a discovery made possible only because the long reads could navigate the highly repetitive genomic background surrounding the enhancer.
Untangling the Chaos of Cancer Genomes
Cancer genomes are notoriously unstable, characterized by extreme structural chaos, copy number variations (CNVs), and the formation of extrachromosomal DNA (ecDNA). Early attempts to map this complexity relied on specialized short-read strategies; for example, Pelin Akan et al. (2012) used long-insert library sequencing to bridge distant genomic fragments, capturing large-scale chromosomal stitching and minute chromosomes in tumor cell lines.
However, modern third-generation sequencing provides a direct, high-resolution alternative. As reviewed by Jing Ren et al. (2023), processing long DNA segments without the need for fragment rejoining provides an unprecedented resolution for large-scale genomic gains and losses in aggressive malignancies like esophageal squamous cell carcinoma.
This structural resolution is even more powerful when brought down to the single-cell level. Lei Chang et al. (2023) developed a single-cell multi-omics method (scGTP-seq) leveraging Nanopore and PacBio long reads. Because ecDNAs and large SVs are highly complex and heterogeneous across individual tumor cells, the long read lengths are essential to accurately identify these structures and directly link them to cell-specific gene expression changes.
When tumor samples suffer from low purity or highly intricate sub-clonal architectures, sequencing can be supplemented by alternative long-range physical mapping. Le Li et al. (2024) demonstrated that optical DNA mapping, which preserves ultra-long continuous DNA molecules, acts as a powerful tool for resolving massive structural rearrangements that defy standard short-read assembly methods.
Tracking Mobile Elements and Somatic Mosaicism
Genomes are not static blueprints; they change over time within individual tissues. This phenomenon, known as somatic mosaicism, is heavily driven by transposable elements—"jumping genes" that copy and paste themselves into new genomic locations. Because transposons consist of highly repetitive sequences, short reads cannot determine their exact insertion sites.
Long-read sequencing solves this by capturing both the repetitive transposon sequence and its unique flanking genomic DNA in a single read. This capability is vital for studying neurodegeneration. Zuguang Li et al. (2025) highlighted how long-read technologies expose complex SVs and retrotransposon insertions accumulating in non-dividing brain cells, linking this somatic mosaicism directly to the progressive pathology of Alzheimer's disease.
To systematically parse these events, computational frameworks must adapt to long-read data. Kevin Riehl et al. (2022) developed TransposonUltimate, a software suite designed to classify and annotate transposons. The authors demonstrated that accurately identifying large structural variants via long reads serves as the necessary foundation for pinpointing precise transposition events across the genome.
Population Benchmarks and Reference-Free Metagenomics
As long-read sequencing matures, the field is moving away from simply identifying variants in isolated individuals toward establishing population-scale benchmarks and reference-free environmental mapping.
Historically, SV detection has suffered from reference bias: if an individual's structural arrangement differs drastically from the standard reference genome, short reads will fail to align properly. To address this in diverse populations, Shuai Sun et al. (2022) integrated multiple long-read sequencing platforms with a trio-binning approach to construct a high-confidence, haplotype-resolved SV benchmark for Asian genomes, significantly reducing false-positive and false-negative rates.
Finally, the benefits of long reads extend beyond single organisms into complex microbial communities. In metagenomics, tracking how structural variants shift across uncultivated microbes is incredibly difficult because most species lack reference genomes. Kristen D. Curry et al. (2024) bypassed this limitation by using long reads to build reference-free co-assembly graphs. Because the long reads natively resolve complex genomic regions and span large structural alterations, the authors could track the abundance and evolution of substantial structural variants directly within mixed environmental samples.
Key Foundational Work
Structural variations in a non-coding region at 1q32.1 are responsible for the NYS7 locus in two large families (2020) — contributed 2 concept elements to this direction.
Comprehensive analysis of the genome transcriptome and proteome landscapes of three tumor cell lines (2012) — contributed 1 concept elements to this direction.