From Genomes to Epigenomes: The Expanding Role of Nanopore Sequencing in Diagnostics

For nearly two decades, microarrays and short-read next-generation sequencing (NGS) have been the workhorses of clinical genomics. SNP and methylation arrays remain cost-effective for genotyping and epigenetic profiling, while Illumina’s Sequencing-by-Synthesis and similar methodologies have similarly transformed diagnostics with targeted panels, whole-exome sequencing, and RNA-seq (Mardis, 2017). These families of technologies have underpinned diagnostics in oncology, rare disease, non-invasive prenatal testing, and infectious disease.

Yet important limitations persist. Short reads struggle to resolve large structural variants (SVs) and repetitive regions; they provide limited haplotype information; and they often require PCR that can bias quantitative signals. Bisulfite treatments for detecting DNA methylation damage DNA and further hinder ability to phase. Array designs constrain what information is gathered and provide limited snapshots of the genomes or epigenomes. Conventional RNA-seq depends on reverse transcription and amplification, which can distort isoform quantification and erase RNA modifications. Crucially, clinical turnaround times can be slow, which is problematic for intraoperative and other time-critical applications.

Third-generation nanopore sequencing, pioneered by Oxford Nanopore Technologies, offers a potential holistic solution to the limitations inherent in microarrays and NGS. Single native DNA or RNA molecules are threaded through nanoscale pores and measured as changes in ionic current, which are decoded to bases and base modifications in real time (Jain et al., 2016; Simpson et al., 2017). Long reads, real-time output, and direct detection of base modifications enable unified assays for SNVs, CNVs, structural variants, and methylation, in a single workflow. Real-time “adaptive sampling” allows for dynamic, selective enrichment of regions of interest without the use of imperfect, potentially biased, hybrid capture.

Importantly, nanopore sequencing is moving from research into practice. Oxford Nanopore and UK Biobank have announced an initiative to generate long-read, and methylation data at population scale, creating a foundational human epigenomic resource (Oxford Nanopore Technologies, 2024). The AI classifier Sturgeon and integrated workflows like ROBIN have demonstrated the feasibility of rapid intraoperative molecular diagnosis for tumours in the central nervous system using nanopore sequencing (Vermeulen et al., 2023; Deacon et al., 2024).

More Comprehensive Variant Detection

The ability to reliably call single nucleotide variants (SNVs) and detect small structural variants, insertions and deletions (INDELs), from nanopore data have improved as base- and variant-calling models mature (Edge and Bansal, 2019; Shafin et al., 2021). Long reads provide phasing, which resolves cis versus trans relationships. This is clinically relevant in compound heterozygosity, pharmacogenomics, and many inherited disorders. The ability to resolve copy number variants (CNVs) and SVs are natural strengths of long read technologies. Rearrangements and breakpoints due to inversions, translocations, and large INDELs can be reliably captured and precisely mapped by individual reads.

Shallow whole-genome nanopore sequencing of plasma cell-free DNA has detected broad copy number aberrations and informative fragmentomic patterns while also preserving methylation information (Martignano et al., 2021).

In neuropathology, CNV profiles complement methylation-based tumour classification. In a landmark Nature study of intraoperative central nervous system tumour diagnosis, shallow nanopore sequencing identified hallmark CNVs such as 1p and 19q co-deletions from as few as 20,000 to 50,000 reads, enabling confident calls in under two hours (Vermeulen et al., 2023).

Single-Molecule Epigenomics

Direct detection of cytosine methylation from native DNA is a unique capability of nanopore sequencing. Unlike bisulfite conversion, which damages DNA and obscures strand information, nanopore sequencing preserves the integrity of DNA molecules and calls methylation directly from the raw electrical signal. This allows haplotype-aware analyses and the exploration of long-range epigenetic patterns that are invisible to short-read approaches (Simpson et al., 2017; Liu et al., 2019).

One particularly informative metric is methylation entropy, which quantifies the variability in CpG status across neighbouring sites. High entropy reflects epigenetic instability, a hallmark of many cancers, while low entropy indicates stable and well-maintained programmes. In liquid biopsy studies, single-molecule entropy profiles derived from nanopore sequencing have been shown to distinguish tumour-derived fragments from background hematopoietic cfDNA, improving the sensitivity of early detection and enabling longitudinal monitoring of treatment response (Lau et al., 2023).

Long reads also make it possible to resolve methylation haplotypes by phasing methylation states across kilobase-scale regions. This enables allele-specific methylation analysis, which is important for imprinting disorders, regulatory variant interpretation, and the study of tumour heterogeneity (Ni et al., 2020). Beyond haplotypes, nanopore reads provide meaningful per-molecule methylation estimates. These capture the proportion of CpGs that are methylated along an individual read, linking methylation signatures directly to single molecules. The distribution of these fractions has diagnostic potential, for example by distinguishing clonally methylated tumour molecules from the heterogeneous background of normal cells.

The clinical value of nanopore-based methylation profiling is already evident. In neuropathology, classifiers trained on sparse methylation data have been applied to intraoperative sequencing runs. The Sturgeon classifier achieved accurate classification of CNS tumours within 40 to 90 minutes, demonstrating how methylation analysis with nanopore sequencing can support real-time clinical decision-making (Vermeulen et al., 2023).

The Transcriptome and Epitranscriptome

Nanopore platforms can read RNA molecules directly, which detects distinct isoforms, splice junctions, quantifies poly(A)-tail length, and common RNA modifications such as N6-methyladenosine. Direct RNA sequencing avoids reverse transcription and PCR artefacts, which improves quantification of isoforms and detection of epitranscriptomic changes (Garalde et al., 2018; Workman et al., 2019).

Clinical applications include infectious disease, cancer transcriptomics, and liquid biopsy. During viral outbreaks, direct RNA sequencing has been used to resolve full-length viral genomes and subgenomic RNAs while capturing host response dynamics (Kim et al., 2020). In oncology, long-read RNA sequencing detects clinically relevant fusions such as BCR-ABL1 and EML4-ALK and resolves complex isoforms missed by short reads (Davidson et al., 2020).

Nanopore also supports epitranscriptomic profiling. Benchmarking studies of RNA modification detection using nanopore data show that subgenomic RNAs often carry distinct modification loads that may regulate stability or translation (Maestri et al., 2024). In plasma, early studies indicate that nanopore profiling of cell-free RNA can recover full-length transcripts and novel isoforms in pre-cancerous and cancer cohorts, expanding biomarker potential.

Liquid Biopsy Multi-Modal Readout

Liquid biopsy is a minimally invasive diagnostic approach that analyses molecular material shed into body fluids, most commonly plasma but also cerebrospinal fluid, urine, or saliva. Instead of relying on tissue biopsies, liquid biopsy approaches target cell-free DNA (cfDNA), circulating tumor DNA (ctDNA), cell-free RNA (cfRNA), circulating tumor cells, or exosomes, enabling comprehensive molecular profiling of disease (Wan et al., 2017; Heitzer et al., 2019).

Nanopore sequencing is well suited to this domain because it preserves native molecules and can generate multi-modal readouts in a single assay. From a single DNA library, it is possible to obtain SNVs, SVs, CNVs, phased-methylation haplotypes and patterns of methylation entropy, as well as fragment size distributions and end motifs.

In cancer patients, nanopore cfDNA sequencing has recapitulated copy number aberration profiles and fragmentomic features comparable to short-read methods while adding methylation information that is difficult to obtain without bisulfite conversion (Martignano et al., 2021). Single-molecule methylation analyses separate tumour-derived cfDNA from immune-derived cfDNA and support treatment monitoring with quantitative metrics such as entropy and alpha values (Lau et al., 2023). Addissouky et al. (2024) demonstrated that nanopore sequencing of plasma and cerebrospinal fluid could recover mutations and methylation features of brain tumours, offering a non-invasive diagnostic tool.

Real-Time Targeting and Intraoperative Diagnoses

Adaptive sampling is a software-driven method for selective sequencing. Early sequence from a molecule is basecalled and mapped; if the read does not match target regions, it can be ejected; if it matches, sequencing continues. This enriches target panels or depletes unwanted molecules without laboratory capture and preserves DNA modifications (Loose et al., 2016; Payne et al., 2020).

ROBIN (Rapid nanpOre Brain intraoperatIve classificatioN) integrates adaptive sampling with real-time basecalling, methylation classification, and downstream variant analysis for intraoperative brain tumour diagnostics. In a prospective clinical series, ROBIN reported same-day methylome classifications and next-day integrated molecular profiles with about 90 percent concordance to final diagnoses. Adaptive sampling increased CpG yield and classifier confidence (Deacon et al., 2024). Combined with Sturgeon, which is trained on sparse CpG data, these approaches exemplify nanopore’s potential for rapid intraoperative diagnostics.

Population-Scale Long-Read Epigenomics

To translate long-read and epigenetic assays into population health, large reference datasets are essential. The Oxford Nanopore and UK Biobank collaboration aims to generate long-read genomes with methylation detection for 50,000 participants, creating the first epigenomic reference at this scale (Oxford Nanopore Technologies, 2024).

For diagnostics, this resource will benchmark methylation entropy, haplotypes, and alpha values across ancestry groups and life stages; it will support training and validation of classifiers such as Sturgeon; and it will facilitate discovery of integrated genomic and epigenomic biomarkers when combined with UK Biobank proteomic, imaging, and outcome data (Bycroft et al., 2018).

Challenges and Outlook

Challenges remain before widespread clinical adoption. Sensitivity in low allele fraction settings such as plasma cfDNA requires sufficient depth, robust error suppression, and careful molecule-level analysis. Although nanopore accuracy continues to improve, very low frequency variant detection requires stringent validation and orthogonal confirmation (Wenger et al., 2019).

Computational demands can be significant however. Real-time basecalling, methylation calling, per-molecule modification computation, haplotype phasing, and adaptive sampling require robust pipelines and hardware acceleration. Clinical validation demands multi-center studies that demonstrate reproducibility, specificity, and impactfulness. Costs and sample preparation must be optimised for routine use in the clinic.

Despite these hurdles, chemistry and basecalling continue to improve, population-scale resources are arriving, and clinical pilots are demonstrating feasibility. Together, these trends indicate that nanopore sequencing is on a trajectory toward mainstream use as an integrated, multi-modal diagnostic platform.

Nanopore sequencing brings together what have long been separate strands in clinical genomics. Early clinical demonstrations such as Sturgeon and ROBIN, combined with population-scale efforts like the UK Biobank collaboration, point to a near future where diagnostics are more informative, real-time, and integrated.

Author - Jack Monahan, Senior Scientist at Hurdle