Cancer liquid biopsies by Oxford Nanopore Technologies sequencing of cell-free DNA: from basic research to clinical applications

Abstract

Liquid biopsies, in particular, analysis of cell-free DNA, are expected to revolutionize the current landscape of cancer diagnostics and treatment. However, the existing methods for cfDNA-based liquid biopsies for cancer have certain limitations, such as fragment interruption and GC bias, which are likely to be resolved by the emerging Oxford Nanopore Technologies (ONT), characterized by long read-length, fast read-times, high throughput, and polymerase chain reaction-free. In this review, we summarized the current literatures regarding the feasibility and applications of cfDNA-based liquid biopsies using ONT for cancer management, a possible game-changer that we believe is promising in detecting multimodal biomarkers and can be applied in a wide range of oncology utilities including early screening, diagnosis, and treatment monitoring.

Graphical abstract

Background

Cancer imposes a significant impact on human health worldwide, with approximately 10 million people dying from cancer each year [1]. Although tissue biopsy is the gold standard for cancer diagnosis [2], it is limited by constraints on invasive procedures, poor patient compliance, and sampling frequency [2, 3]. CfDNA, a crucial component of liquid biopsy biomarkers, has emerged as a promising noninvasive alternative to tissue biopsy in cancer diagnosis and prognosis prediction [4]. CfDNA fragments are shed into body fluids by dying cells during apoptosis, necrosis, or secretion [5]. The fragments exhibit a wide size distribution, ranging from 20 bp to over 1 kb, with a notable peak at 166 bp. They predominantly circulate in a linear double-stranded configuration, also existing as linear single-stranded and circular forms [5, 6]. In addition, the half-life is variable ranging from 15 min to several hours [7]. Of note, the pool of cfDNA molecules contains rich molecular features, including those involved in genetics (e.g. mutation, copy-number variation (CNV), structural variation (SV)), epigenetics (e.g. methylation and other modifications) [5], and fragmentomics (e.g. fragment size, coverage, and end motif) [4]. When tumors develop, progress, and disseminate, changes in cfDNA profiles occur. Analyzing these features can thus be used for cancer detection, molecular subtyping, surveillance, and prognostication [8,9,10].

Currently, PCR-based and next-generation sequencing (NGS)-based methods represent the state-of-art for detection and analysis of cfDNA [11, 12]. Among these, PCR-based methods are the earliest of being applied to cfDNA detection [13]. Despite their affordability, rapidity, and ease of operation, they are limited in screening for known mutations and methylation alterations. NGS, the high throughput sequencing technology, offers opportunities to identify unknown alterations [11, 14]. However, they have limitations in terms of GC bias and loss of methylation information, among others [15]. Recently, new methods have been tested for cfDNA analyses. For example, the Pacbio [16], a long-read single molecule platform, is expected to be capable of resolving the limitations above. However, this single molecule real-time (SMRT) sequencing is less suitable for cfDNA detection, due to the low levels of cfDNA fragments in body fluid and high cost. Thus, it remains an unmet need to develop novel approaches for cfDNA detection and analysis.

Nanopore sequencing by ONT [17], with the characteristics of long read length, fast read times, high throughput [18], and label-free, represents the so-called fourth-generation sequencing technology. This innovative technique has recently emerged for cfDNA detection, particularly in cancer research, yielding unexpectedly promising results [19,20,21,22].

Here, we provide a comprehensive overview of the latest advances of cfDNA in oncology by ONT. We start with the history of cfDNA detection technology and make a comparison among these technologies. Subsequently, we demonstrate the feasibility of ONT for cfDNA detection in cancer research. Then, we focus on the clinical applications of the use of ONT for cfDNA detection, and finally, we discuss the existing limitations and future directions.

Techniques developments and comparison among cfDNA detection methods

Classical methods for cfDNA analysis: PCR-based methods

PCR-based methods are the earliest cfDNA detection methods [13, 23,24,25] and currently, they are still widely used (Fig. 1) [11]. PCR-based methods for cfDNA analysis can be divided into three major categories: real-time quantitative PCR (qPCR), digital PCR (dPCR), and the mass-spectrometry-based method [11]. They are often used for detection of mutations and methylation alterations [26, 27]. Although they are straightforward [28], sensitive and inexpensive [29], they are limited to screening known loci or regions and are unable to detect complex genomic rearrangements or large structural variations. Moreover, the throughput and speed are constrained, particularly in large-scale tumor screening efforts. Additionally, they are not well-suited for discovering novel biomarkers. These significantly hinder their broader applications, especially in the era of precision medicine.

Fig. 1

The historical evolution of cfDNA detection techniques. Overview of the evolution of technologies related with cfDNA detection, spanning from early PCR techniques to NGS methods, and the emergence of nanopore sequencing technologies, highlighting key detection methods at each stage

High-throughput methods for cfDNA analysis: NGS-based methods

The advent of NGS has revolutionized the landscape of cfDNA detection [29]. This high-throughput sequencing methodology has not only expanded the repertoire of cfDNA detection techniques but also refined them, spanning from targeted to non-targeted sequencing, as well as methylation sequencing [11]. NGS has effectively eclipsed the limitations imposed by PCR-based assays, which were historically confined to the low-throughput analysis of specific genetic loci. NGS enables comprehensive high-throughput sequencing across the genomic expanse, thereby facilitating a paradigmatic shift from singular biological perspectives to integrated multi-omics interrogations. The utilization of unique molecular identifiers (UMIs) [30], alongside integration with strategies that enhance sequencing depth and signal-to-noise ratio [31], has markedly improved the sensitivity and specificity of detection assays. Despite these advancements, inherent technological limitations such as PCR amplification leading to fragment interruption and GC-content biases [32, 33], bisulfite conversion resulting in DNA damage [34], lack of method to generate a comprehensive multi-omics dataset in a single sequencing run, remain to be resolved.

Specifically, in terms of targeted sequencing, NGS has demonstrated superior throughput and sensitivity compared to PCR-based methods [35,36,37,38]. For example, Tagged-Amplicon deep sequencing (TAm-seq) [36] has achieved mutation detection sensitivities exceeding 97%, while the Safe-Sequencing System (Safe-SeqS) [37] and Cancer personalized profiling by deep sequencing (CAPP-Seq) [38] have pushed the limits of sensitivity to approximately 98% and nearly 100%, respectively. Despite their high sensitivity, these methods are constrained to the detection of known variations, limiting their utility in informing on novel or unexpected genomic alterations.

NGS has unlocked the potential of untargeted cfDNA sequencing, facilitating a more panoramic perspective on the genome. Whole-genome sequencing (WGS) provides a comprehensive sequencing approach that encompasses both coding and non-coding regions of the genome, thereby increasing the probability of identifying genomic alterations [30]. Lower-depth WGS is currently the most popular method for its cost-effectiveness and has demonstrated effectiveness for identifying mutations and characterizing cfDNA biology [39]. Whole-exome sequencing (WES), an alternative to WGS, is favored for its low input requirements, shorter turnaround time, and high yield [33, 34]. Nonetheless, WES is limited to the analysis of the protein-coding genome, which precludes the identification of mutations in non-coding regions. This limitation can lead to a loss of critical information. For example, WES fails to detect mutations at the transcription start site (TSS) within the promoter region, a lacuna that could impede researchers from uncovering the influence of the entire transcription processing.

Sequencing-based methods for methylome analysis can be classified into two groups: bisulfite conversion–based and enrichment-based methods [40]. The former is considered to be the gold standard of DNA methylation analysis [41], however, they are hampered by inefficiencies due to the inability to differentiate between 5mC and 5hmC [42], the degradation of input DNA during conversion, high costs, and limited coverage. Enrichment-based methods [43, 44], while potentially avoiding DNA damage associated with bisulfite conversion and enriching for CpG-rich regions, are restricted to the analysis of methylated cytosines in CpG islands and necessitate a significant input of cfDNA. This requirement may pose a challenge when working with samples where cfDNA is present in limited quantities.

As the innovative paradigm of cfDNA multi-omics analysis gains traction, there is a burgeoning interest among researchers to integrate analyses of fragmentomics, methylomics, and genomics. Nonetheless, NGS-based methods face intrinsic limitations in concurrently generating multi-omics datasets within a single sequencing event, at most yielding methylomics and fragmentomics features in one go. This constraint impedes the progression of research endeavors that require a holistic and simultaneous interrogation of multiple layers of cfDNA.

Emerging methods for cfDNA analysis: ONT-based method

Oxford Nanopore Technologies, initially conceptualized in the laboratories of David Deamer, George Church, and Hagan Bayley during the 1980s [45, 46], has undergone maturity through the optimization of nanopores and motor proteins, multiple times of sequencing, and the development of novel base-calling algorithms, that ultimately leads to its commercialization in 2015. It relies on nanometer-scale protein pores, known as “reader proteins” [47], to serve as biological sensors immersed within an electrolytic medium. As DNA strands traverse the nanopore, they elicit distinct electrical signals for each base, with epigenetic modifications on these bases also manifesting in the current signals. Engineered hardware components, specifically designed circuits for signal processing, in conjunction with base-calling algorithms and software, transform the nanopore readings of both canonical and modified bases into definitive sequence information [48,49,50]. This sequencing technology, characterized by its straightforward detection principle, portability [51], long read-length [52], fast read-times [18], label-free, and high-throughput properties, is on par with PacBio and constitutes a significant part of single-molecule long-read sequencing platforms [53]. Although PacBio can also be used for cfDNA detection, ONT could offer additional advantages such as shorter turnaround time, lower cost, and lower input load. A summary table highlighting the differences between PacBio and ONT for cfDNA sequencing is provided as Table 1.

Table 1 PacBio and ONT for cfDNA sequencing

The success of ONT-based DNA sequencing has inspired research into the sequencing of cfDNA (Fig. 2). It provides an integrated analytical framework that encompasses a spectrum of molecular insights, encompassing genomic, epigenetic, and fragmentomic layers simultaneously. This not only enhances analytical precision and reduces intra-modal bias, but also significantly amplifies the probability of detecting cfDNA. Moreover, it streamlines costs and mitigates batch effects. Furthermore, low-coverage ONT sequencing analyses have emerged as a cost-effective strategy that sustains the momentum of research endeavors, while maintaining robustness in cfDNA detection capabilities. However, the complexity of the data generated by these advanced techniques presents new challenges. The discernment of pertinent information from the deluge of multi-omics data necessitates sophisticated bioinformatics strategies. The interpretation of inter-omics exclusivity, concordance, and the reconciliation of quantitative disparities between distinct omics layers remain formidable analytical hurdles.

Fig. 2

Comparison of cfDNA sequencing technologies. A comprehensive comparison of cfDNA sequencing technologies crossing third generations has been conducted based on detection principles, read length, runtime, and target applications, with a detailed list of their advantages and disadvantages

Leveraging the characteristics of long-read sequencing, ONT enables uninterrupted sequencing across repetitive regions or other complex genomic structures, which is particularly advantageous for the detection of structural variations (SVs) in the genome.

Additionally, by eliminating the procedure of PCR amplification and bisulfite treatment, ONT has the potential to minimize the loss of DNA molecular information, which is a critical consideration in the preservation of epigenetic signatures. It is noteworthy that nanopore sequencing can distinguish between 5mC and 5hmc without additional processing. Furthermore, the technology can also detect 6 mA [56], acting as an important epigenetic modification implicated in the promotion of tumorigenesis and cancer progression [57]. Despite the implications of 6 mA, research in this domain has been constrained by the limitations of previous detection methodologies.

The technology offers other advantages, including high throughput, rapid detection times, and portable equipment, all of which are conducive to the discovery of cfDNA biomarkers and fundamental research [16]. Although the low read accuracy of ONT remains a limitation, it is anticipated that with ongoing optimization of nanopores and motor proteins, the implementation of multiplexed sequencing, and the refinement of base-calling algorithms, the accuracy will be substantially improved [58,59,60].

Applications of cell-free DNA in Oncology by ONT

CfDNA detection plays a pivotal role across multiple fields, including pregnancy [61], cancer [62], transplantation [63], and autoimmune diseases. In oncology, cfDNA offers significant value in tasks including diagnosis, treatment monitoring, therapeutic assessment, and prognosis [64,65,66]. However, as previously mentioned, current detection methodologies are approaching their sensitivity thresholds [67], which limits their broader clinical applications. ONT has emerged as a new paradigm in cfDNA detection, potentially addressing the limitations of traditional methods and offering enhanced potential for both research and clinical applications. Its ability to provide long-read sequencing, real-time analysis, and detection of multiple molecular features positions it as a promising tool in the field of oncology. In the following part, we summarize basic research and clinical applications of cfDNA analysis by ONT, covering a range of genomic alterations, methylation patterns, and fragmentation analysis, focusing on issues associated with cancer management (Fig. 3). A detailed list of relevant literature is provided as Table 2, which serves as a reference for the subsequent in-depth discussion on the multifaceted applications of ONT in cfDNA-based oncology research and practice.

Fig. 3

Clinical applications of cell-free DNA profiling in Oncology by ONT. a CfDNA isolated from a patient’s blood sample and sequenced by ONT. b Then subjected to comprehensive profiling to assess fragmentation patterns, mutations, methylation status (Me), copy number variations, and chromosomal rearrangements. c-d These profiling data analyzed, either individually or in combination. e And applied to cancer diagnosis, treatment monitoring, mutation tracking, and tumor classification

Table 2 Summary of cfDNA-based studies by ONT in oncology

Studies involving genetic alterations

In essence, all cancer arise as a result of genetic alterations [76]. Under the influence of diverse internal and external carcinogenic factors, cells undergo genetic mutations that lead to abnormal gene structures and dysregulated gene expression [77]. These changes significantly impact the biological activity and genetic characteristics of the cells, resulting in tumor cells that differ from normal cells in morphology, metabolism, and function. This process is driven by multi-gene and multi-step mutations [78]. Throughout the initiation and progression of tumors, continuous genetic changes occur, making the study of genomic alterations essential for both basic cancer research and clinical applications.

Mutation-based studies

Genomic alterations can be broadly categorized by fragment length into short-segment mutations (< 50 bp) and structural variations (> 50 bp). The latter include single nucleotide variants (SNVs) and small insertions or deletions (Indels) [79, 80]. These are the most common genomic changes observed in tumors and play a pivotal role in driving cancer initiation, progression, and metastasis. Recent studies have increasingly leveraged ONT to detect these mutations, with clinical correlation with clinical cancer diagnostics, treatment monitoring, etc. For example, Burck et al. detected ERBB2 F310S and PIK3CA H1047R mutations in plasmids and mice plasma [73]. They provided a novel approach to distinguish between wild-type and mutant variants, opening new pathways for early cancer diagnosis.

In addition to direct sequencing of cfDNA from body fluids using nanopore technology, researchers have also applied it to PCR-amplified fluid samples. Bruzek et al. employed PCR-amplified cf-tDNA to identify genetic mutations in the cerebrospinal fluid (CSF) of pediatric high-grade glioma (p-HGG) patients, achieving a sensitivity of 85% and a specificity of 100%. They further performed serial sequencing of CSF cf-tDNA and successfully monitored multi-gene molecular responses, demonstrating the utility of continuous molecular surveillance in clinical trials [71]. CyclomicsSeq, a method that involves rolling circle amplification (RCA) to amplify the TP53 target gene, followed by nanopore sequencing, successfully applied to monitor tumor burden in head and neck cancer patients during treatment [72]. Another advancement, is the combination of blocker displacement amplification (BDA) with ONT to detect EGFR mutations in the plasma of patients with non-small cell lung cancer (NSCLC) [81]. This approach provided a reliable method for clinical detection of EGFR mutations, which are crucial for guiding targeted therapies.

Although incorporating PCR-based methods may help reduce sequencing errors and improve the distinction of true mutations, it comes at the cost of losing the primary advantage of using unamplified cfDNA as well as introducing PCR-related errors or duplications. Other approaches are also available to address the possible sequencing errors, such as integration with error correction strategies or conversion to NGS-liked short reads for downstream analysis of mutation detection [82,83,84]. Additionally, ONT can also be used for analysis of collective mutational features that are much tolerant to base-errors, such as mutational signatures that have been shown as useful for cfDNA-based cancer detection [85, 86].

Structure variations-based studies

Structural variations represent another important class of genomic alterations. Unlike SNVs, SVs are more complex and variable, affecting a larger portion of the cancer genome than any other type of somatic genetic alteration [87]. They range in size from ~ 50 bp to well over several megabases of sequence. SVs encompass numerous subclasses, including unbalanced copy number variants, deletions, insertions, chromosomal gains and losses, as well as complex DNA rearrangements [80].

Currently, nanopore technology focuses on detecting cfDNA structural variants in CNVs and chromosomal rearrangements. Martignano et al. detected common pathogenic CNVs in the plasma of NSCLC patients, such as MYC and PIK3CA amplifications [68]. Afflerbach et al. analyzed CNV profiles for the classification and identification of prognostic markers for brain tumors, such as C19MC amplification, chromosome 1q gain, and chromosome 6q loss [70]. Tumor fraction (TF), which refers to the proportion of circulating tumor DNA (ctDNA) within the total cfDNA, is a key parameter in cfDNA analysis [88]. It can be effectively detected using ichorCNA via ONT [20, 21], providing valuable insights into tumor burden and disease progression.

VDJ rearrangements are a kind of chromosomal rearrangement, which often occurs in patients with hematological malignancies [89]. Sampathi et al. developed a workflow based on nanopore sequencing of PCR-amplified VDJ rearrangements in the cfDNA from bone marrow biopsy samples, providing a way to assess B-cell acute lymphoblastic leukemia (B-ALL) heterogeneity and monitor disease progression [75], which suggests its potential as a valuable complement to current cell-based clinical assays.

These studies illustrate the potential of ONT sequencing for detecting a wide range of genomic changes in cfDNA and highlight its utility across various cancer types; However, aside from the technical challenge of improving the accuracy of mutation detection, the additional complexity of translating the results into clinical practice remains a significant barrier. The dynamic nature of ctDNA and the heterogeneity of tumors require comprehensive longitudinal studies to validate the clinical applicability of these technologies.

Methylation-based studies

Tumorigenesis is closely linked to alterations in DNA methylation, with changes in methylation patterns typically occurring more frequently and at earlier stages than genetic mutations. These epigenetic modifications exhibit remarkable stability and cell-type specificity, enhancing the potential to detect tumor-related signals, even during the early stages of cancer development [90, 91]. Leveraging the unique advantages of Oxford Nanopore Technologies for cfDNA methylation detection, numerous high-performance studies have emerged, underscoring its value in clinical utilization.

One of the key hallmarks of the cancer epigenome is global DNA hypomethylation, which can lead to nuclear disorganization and the loss of gene silencing in genes associated with cellular proliferation [90, 92]. This phenomenon has long been proposed as a general marker for ctDNA. Recently, Katsman et al. further validated it by observing global hypomethylation in 5 out of 6 non-small cell lung cancer plasma samples [20].

Beyond its involvement in tumorigenesis, DNA methylation plays a critical role in cellular differentiation [93], providing a pathway to trace the origin of circulating free DNA to its primary cells or tissues [94,95,96]. This capability is particularly important for assessing treatment-related toxicity and enhancing our understanding of cancer biology. In a study analyzing plasma cfDNA from lung cancer patients and healthy individuals, researchers successfully deconvoluted cfDNA methylation patterns using both ONT and NGS [20]. The results obtained from these two sequencing platforms were highly consistent, enabling precise estimation of the cellular origins of cfDNA from both healthy and cancerous plasma. This study further confirmed the efficacy and reliability of ONT in cfDNA deconvolution. Following this, Yu et al. employed an alternative method that leveraged cell type-specific methylation patterns to trace the tissue origin of cfDNA [22]. Using the results of this tissue-of-origin analysis, they successfully distinguished HCC patients from HBV carriers with 100% accuracy, significantly surpassing the 75% accuracy achieved with PacBio sequencing. Similarly, this technique was successfully applied to monitoring colorectal cancer patients, yielding highly satisfactory results [69].

Brain tumor diagnosis has historically posed significant challenges due to the difficulty in obtaining tissue samples and the complexity of the diagnostic process, whereas recent advances in DNA methylation analysis have provided substantial benefits [6]. Afflerbach et al. recently applied ONT to analyze cfDNA methylation in cerebrospinal fluid from brain tumor patients, successfully classifying tumors with high concordance to histopathological diagnoses [70]. This further supports the use of methylation profiling for the accurate classification of brain tumors.

Collectively, these studies demonstrate that the nanopore sequencing platform can be effectively employed to determine the tissue-of-origin of cfDNA, distinguish between cancerous and non-cancerous states, and classify cancer subtypes. While these findings are promising, further clinical applications and studies are necessary to expand ONT’s use in cancer detection, particularly in the areas of early cancer screening and personalized medicine.

Fragmentation-based studies

Cancer fragmentomics is an emerging field focused on studying the molecular and associated properties of cfDNA fragments in cancer patients and non-cancer individuals, aiming to understand and leverage the differences in fragmentation patterns between tumor-derived cfDNA and normal cfDNA for cancer research [97,98,99].

CfDNA is a mixture of fragmented DNA molecules released from different tissues within the body. The size and distribution of these fragments show significant differences between healthy individuals and cancer patients. Increasing evidence suggests that ONT can effectively distinguish cancer patients from healthy individuals based on cfDNA fragment distribution. Examples include the use of plasma-derived cfDNA to differentiate lung cancer patients [20], urinary cfDNA to distinguish bladder cancer patients [21], and cerebrospinal fluid cfDNA to distinguish brain tumor patients [71]. One of the key advantages of ONT sequencing is its ability to detect long cfDNA fragments. In a study focused on bladder cancer patients, ONT sequencing revealed that the median proportion of cfDNA fragments longer than 300 bp was 62.4%, compared to only 14.2% detected by short read sequencing [21]. Additionally, ONT was able to detect the longest cfDNA fragments, reaching up to 4.28 kbp. This suggests that ONT is more capable of accurately reflecting the true length distribution of cfDNA.

End motifs represent a common feature in cfDNA fragmentomics. Initial studies suggested that end motifs could serve as standalone biomarkers for distinguishing cancer patients from healthy individuals [100]. However, recent research has revealed that end motifs are highly sensitive to variations in library preparation and sequencing platforms, challenging their reliability as independent diagnostic indicators. Specifically, Katsman et al. analyzed the end motifs of plasma cfDNA from lung cancer patients using both ONT and Illumina sequencing platforms [20]. The results showed discrepancies in the frequency of detected end motifs between the two platforms. Most low-frequency motifs appear at slightly higher frequencies with Nanopore sequencing, while high-frequency motifs appear less frequently in Nanopore. Among them, the “CCCA” end motif shows the most significant frequency disparity between the two platforms, both in healthy individuals and cancer patients, suggesting that end motifs may be particularly sensitive to changes in library preparation strategies and sequencing platforms. ONT technology has been shown to have a lower 3’ end error rate as compared with Illumina NGS [101]. Nevertheless, the diagnostic accuracy of ONT-based end motif analysis for cancer detection remains to be examined in future studies, and it could be interesting to further explore the performance by integrating end motif with other fragmentomics features, like previously did by us and other teams with NGS data [102].

Coverage referring to the distribution of cfDNA fragments across the genome during sequencing, is crucial for foundational cfDNA research [103]. Recent studies have demonstrated that ONT excels at capturing coverage information near key regulatory elements, including TSS, nucleosome-depleted regions (NRR), and CCCTC-binding factor (CTCF) binding sites [20, 21].

While nanopore technology has demonstrated significant advantages in studying cfDNA fragmentomics, current research remains in its early stages. Future studies should focus on incorporating additional fragment characteristics, such as integrated fragmentation score (IFS), window protection scores (WPS), and orientation-aware cell-free fragmentation (OCF), to conduct more in-depth investigations of cfDNA dynamics.

Multimodal analysis‐based studies

Traditionally, research on cfDNA in cancer has primarily relied on single-omics approaches, such as mutations or methylation alternations [104]. However, cfDNA exhibits significant variability and heterogeneity, not only among different patients but also within the same patient at different times of the day. Moreover, the proportion of tumor-derived DNA fragments in circulation is typically low, thereby making single-modal approaches less reliable [105].

Multi-modal analysis, which integrates two or more modes of data for comprehensive analysis, has emerged as a promising approach in cfDNA cancer detection [31, 39]. By combining multi-dimensional features, more comprehensive and complementary information can be obtained, potentially enhancing the accuracy and efficiency of diagnosis and treatment. For example, Afflerbach et al. applied both CNV and methylation profiling to classify brain tumors [70]. CtDNA was successfully detected in 50 out of 129 samples, while only 5/50 samples with detected ctDNA contained tumor cells detectable through microscopy. 32% (n = 16/50) of cases were correctly classified by both CNV and methylation profiling. CNV analysis detected ctDNA in 56% (28/50) of cases, while methylation profiling identified ctDNA in an additional 12% (6/50) of cases. Subsequently, ichorCNA analysis of CNV profiles revealed diagnostic and prognostic biomarkers, such as C19MC amplifications in embryonal tumors with multilayered rosettes or Chr.1q gains and Chr.6q losses in posterior fossa group A ependymoma. Multi-modal-based assays have also been demonstrated to improve the specificity and sensitivity of cancer diagnosis, as well as the molecular subtyping of cancer, but have not yet been evaluated by ONT.

Despite the promise of multi-modal approaches, the generation of large, complex datasets introduces several challenges. While the advantages of multi-omics integration have been demonstrated, questions remain regarding how to handle the specificity of different features, such as the tissue-specific signals of methylation patterns, the critical role of mutations in cancer progression, and the distinct fragmentation profiles between cancer and non-cancer patients. When integrating these diverse omics data, issues such as feature exclusivity and how to assign appropriate weights to different modalities arise. Additionally, there is a need for further research into how to analyze and interpret the vast amounts of complex data generated by these methods. As research in this field progresses, addressing these challenges will be crucial in realizing the full potential of cfDNA-based multi-modal analysis for cancer diagnosis and prognosis.

Future directions and opportunities

Improving cfDNA enrichment efficiency

Although nanopore sequencing technology can be combined with PCR to compensate for the limited quantity of cfDNA, this approach brings side-effect of erasing epigenetic modifications such as methylation and hydroxymethylation, as well as possible changes to fragment-length-based characteristics. Therefore, to fully exploit the advantages of nanopore sequencing for unamplified cfDNA, a sufficient amount of cfDNA (at least 25 ng) remains a prerequisite for accurate analysis. However, in practical applications, obtaining enough cfDNA remains a practical and significant challenge [50]. The inherent limitations of cfDNA represents the primary obstacle. The concentration of cfDNA in plasma is extremely low [34], with only a few thousand cfDNA molecules per milliliter of plasma [67]. Furthermore, cfDNA has an exceptionally short half-life [106], typically less than 2 h, which can lead to degradation before the extraction process is even completed. In addition to these biological constraints, external factors introduced during the extraction process, such as oxidative or deamination damage, contamination by cellular genomic DNA [29], and artifacts during human operations, may further exacerbate cfDNA loss and degradation, significantly reducing the amount of cfDNA available for subsequent procedures [107]. Therefore, improving the efficiency of cfDNA extraction has become a pressing issue.

We propose that the first step should be to standardize the cfDNA extraction process to avoid human-induced cfDNA loss. This includes implementing standardized sample handling procedures, ensuring the use of high-quality cfDNA extraction kits, and optimizing post-extraction storage conditions to minimize sample loss and enhance cfDNA extraction efficiency. In addition to optimizing existing biological extraction methods, new cfDNA enrichment techniques based on chemical and physical approaches should be explored. For example, metal–organic framework (MOF)-based cfDNA enrichment and microfluidic chip-based cfDNA enrichment are emerging as promising areas of research [108, 109]. These techniques leverage physical or chemical interactions between DNA and various materials, offering new routes for cfDNA enrichment and potentially significantly improving the efficiency of cfDNA recovery. Although these methods are still in the research phase, they provide innovative solutions to the problem of insufficient cfDNA enrichment.

Developing appropriate algorithms

While cfDNA detection using ONT has shown some progress in cancer research, the limitation in cfDNA quantity often results in low-depth sequencing, with average coverage typically below 1X. This low coverage means that certain tumor-related signals present in cfDNA may not be accurately captured, thus compromising clinical decision-making. For instance, the detection sensitivity of low-frequency mutations is significantly reduced, increasing the risk of false-negative results; The estimation of TF also becomes inaccurate, affecting the assessment of a patient’s tumor burden. Therefore, developing algorithms capable of accurately identifying cfDNA features under low sequencing depth conditions is of critical importance. ONT-based cfDNA detection can capture not only genomic information but also epigenomic and fragmentomic data. Theoretically, integrating these layers of omics can yield more accurate results. However, most current algorithms are designed for single-omics analysis, posing risks of overfitting [110]. Moreover, these algorithms may cause cancer-specific signals from one omic layer to be masked or canceled out by signals from another layer. Thus, developing new algorithms tailored for multi-omics cfDNA analysis is a key task for the future.

In addition to developing new algorithms, more automatic algorithms for cfDNA analysis should be created. For instance, end-to-end machine learning (ML) pipelines can preprocess data, automatically select the appropriate ML model, and determine hyperparameters without any user input [110, 111]. This provides significant convenience for bioinformatics beginners, particularly for clinicians in practical settings. Furthermore, constructing end-to-end ML pipelines allows ML engineers to build once and reuse multiple times, making them highly efficient and suitable for clinical production environments. However, currently, there are very few automated algorithms specifically designed for cfDNA research. Among them, only ichorCNA is widely utilized [88]. Moreover, there is a noticeable lack of pipelines tailored for cfDNA analysis using nanopore sequencing technology, which limits its broader application in cancer research. In the future, as ONT becomes more widely used for cfDNA detection, the integration of appropriate machine learning methods will facilitate tumor information analysis, driving cfDNA toward more widespread clinical applications.

Service for clinical routine

In fact, ONT is a single-molecule sequencing technology, theoretically capable of sequencing a wide range of small molecules, with nucleic acids being the most commonly detected [112,113,114]. Beyond cfDNA, ONT has been explored for the detection of other nucleic acids in cancer research. For example, Luo et al. used ONT to sequence extrachromosomal circular DNA(eccDNA) in plasma and tissue samples after RCA, finding that eccDNA had superior diagnostic performance compared to traditional markers such as CEA and CA-199, demonstrating the potential of eccDNA as a cancer diagnostic marker [115]. Additionally, Reggiardo et al. utilized nanopore technology to detect circulating repetitive RNA in plasma, discovering that each type of cancer had unique circulating RNA signatures [116]. ONT has also been applied to the detection of pathogenic microorganisms’ DNA/RNA, further expanding its utility [117, 118]. Recently, researchers have even attempted to apply ONT to protein detection [19]. Although this approach is still in its infancy due to the complexity of protein folding and the diversity of amino acids, the potential for future applications remains promising.

Recent works have also combined nanopore sequencing with single-cell technologies. For example, Thijssen et al. linked somatic mutations to transcriptional changes at the single-cell level, uncovering a series of previously unidentified splicing interruptions in the blood of chronic leukemia patients, such as BCL2 G101V, and subclonal losses of NOXA and BAX, which were associated with venetoclax resistance [119]. Cortés-López et al. utilized long-read single-cell transcriptomics to identify cell-identity-dependent mis-splicing mediated by SF3B1 mutations. They found that SF3B1mut cells exhibit lineage skewing and stage-specific mis-splicing in myelodysplastic syndrome with SF3B1mut cells showing erythroid lineage bias and mirroring MDS-related mis-splicing in clonal hematopoiesis, thereby revealing the mechanisms of SF3B1 mutations in patients [120]. These reports suggest that ONT technology has enabled single-cell sequencing to overcome previous limitations in mapping full-length RNA, making it possible to quantify different splicing events within the same mRNA transcript. On the other hand, conversely, it is likely that the integration of single-cell technologies would further strengthen the potential and utility of ONT-based cfDNA analysis for cancer detection and classification, particularly in terms of identification and validation of tumor-heterogeneity-aware biomarkers, biological interpretation of machine-learning models using ONT-based cfDNA data, single-cell sequencing-guided transformation and feature generation of ONT-based cfDNA data (e.g., deconvolution to abundance scores of cell populations), as well as the joint modeling of ONT-based cfDNA markers and single-cell analysis derived cellular markers, similar to previous work by us and other teams based on NGS data [121, 122].

From a cancer biology perspective, tumorigenesis and progression are complex processes involving multiple layers, including the genome, transcriptome, and proteome [119, 123]. Integrating multi-layered, multi-angle information for cancer analysis may yield more accurate answers. We hypothesize that by using ONT to study gene structure at the DNA level, investigate gene alterations at the RNA level, and assess gene function at the proteome level, we can achieve the most comprehensive understanding of cancer. Fortunately, ONT technology is capable of accomplishing these tasks simultaneously, requiring only a MinION sequencer and a tube of blood (Fig. 4).

Fig. 4

Future perspectives of applying Nanopore technology for cancer liquid biopsies. a Improving cfDNA extraction efficiency through the standardization of experimental procedures and the development of innovative detection technologies. b Developing new algorithms and establishing automated end-to-end machine learning pipelines to facilitate clinical research. c Leveraging nanopore technology to detect additional biomarkers and integrating with other technologies (e.g., single-cell sequencing), to accelerate its application in early cancer screening, diagnosis, treatment monitoring, and drug resistance detection

Conclusions

CfDNA test technologies are advancing rapidly. As an emerging approach for cfDNA analysis, nanopore sequencing offers several distinct advantages, particularly its capability for direct sequencing without the need for PCR amplification or bisulfite conversion. It has already shown great potential in applications such as analyzing fragmentomics, methylomics, as well as genomic alterations in both single and multi-modal analyses, spanning from basic research to clinical applications. Despite the technology’s current limitations, primarily concerning base-calling accuracy and algorithmic development, it is likely that the continued evolution of compatible algorithms will unlock the full potential of nanopore technology for cfDNA-based liquid biopsies.

Looking ahead, nanopore sequencing is poised to capitalize on its single-molecule sequencing capabilities, enabling the detection of a wide range of biomolecules and integrating with platforms like single-cell sequencing technologies. These advancements will pave the way for more precise, comprehensive, and minimally invasive tools for cancer diagnosis, therapeutic monitoring, and prognostic evaluation, ultimately driving the progress of precision oncology.