2.1 Bisulfite conversion-based methods

Bisulfite conversion-based methods exploit the unique property that cytosine residues can be converted to uracil in genomic DNA, but methylcytosine remains unchanged after bisulfite treatment, facilitating the identification of 5-methylcytosine (5mC) and cytosine.⁵² Techniques such as WGBS, SureSelect Methyl-Seq, and SeqCap Epi CpGiant reduced representative bisulfite sequencing (RRBS)⁵³ (Figure 2B), methylated CpG tandem amplification and sequencing (MCTA-seq), and methylation arrays have been developed.⁵⁴ These methods enable precise identification of DNA methylation changes at the single-base level, though genomic coverage varies. Based on this, Shen et al.⁵⁵ used a loop-mediated isothermal amplification method based on thiol-primed primers to achieve high sensitivity detection of ctDNA E-Box methylation in tumor tissues and biological samples of renal cell carcinoma patients. Therefore, researchers are advised to select the most suitable technique based on the specific requirements of their study.

Bisulfite sequencing emerges as the most dependable technique for DNA methylation analysis, offering precise quantification at the single-base level.⁵⁶ Addressing the challenges posed by bisulfite treatment, such as DNA damage and reduced sequence diversity, Liang et al.⁵⁷ introduced ELSA-seq. WGBS technology facilitates comprehensive identification of methylation sites across the genome.⁶ Our recent advancement in ctDNA–WGBS methodology enables the accurate determination of genome-wide methylation patterns from a minimal volume of ctDNA, extracted from merely 200 µL of plasma, as shown in Figure 3A. This refined approach, utilizing as little as 1 ng of DNA for library preparation, achieves unbiased genome-wide coverage and employs a comprehensive computational strategy to minimize background noise from low-repeat and nontumor-derived ctDNA. This innovation allows for early BC screening across multicenter patient cohorts with high specificity and sensitivity, developing highly precise biomarkers for distinguishing various cancer types and subtypes.⁵⁰ Such advancements hold significant potential for clinical application. GRAIL has harnessed bisulfite sequencing combined with machine learning to develop a methylation pattern classification method, achieving 67% sensitivity and 99% specificity in detecting early-stage disease across multiple cancer types.⁵⁸ Huang et al.⁵⁹ utilized WGBS to profile DNA methylation in small volumes of CSF ctDNA from children with medulloblastoma (MB), demonstrating its utility in pediatric oncology. However, the limitations of bisulfite conversion, particularly its inefficiency in capturing DNA prone to fragmentation during apoptosis,⁶⁰ highlight the importance of methodological consideration in cfDNA methylation studies. Expanding upon this, Maggi and colleagues⁶¹ have developed a method for comprehensive cfDNA methylation analysis, producing high-quality libraries with a 98% conversion rate, suitable for plasma-based studies.

2.2 Nonbisulfite conversion methods

Nonbisulfite conversion methods for DNA methylation analysis include antibody enrichment and restriction endonuclease techniques. The antibody enrichment approach, particularly MeDIP (methylated DNA immunoprecipitation), uses a 5mC antibody to selectively enrich methylated DNA fragments, identifying methylation levels through fluorescence differences.⁶² This category encompasses sequencing methods like MeDIP-Seq and MBD-Seq,⁶³ where the development of recombinant protein complexes with enhanced affinity is crucial. cfMeDIP-seq, for instance, has shown high sensitivity in ctDNA methylation detection, effectively identifying and classifying various cancer types.⁶⁴ Nuzzo et al.⁶⁵ utilized cfMeDIP-seq to detect ctDNA with 97% sensitivity and 100% specificity in a study on 69 individuals with renal carcinoma at different stages. Methyl-sensitive restriction endonucleases, another approach, exhibit varying sensitivities to methylcytosine, aiding in methylation detection.⁵² However, their application is limited by the specificity of enzyme recognition sites, and methylation-specific restriction endonucleases (MSRE) are costly and less effective in regions of intermediate methylation. Recently, Gouda et al.⁶⁶ introduced a sequencing method targeting 32 specific CpG sites for methylation detection in CRC. Before clinical application, these methods necessitate extensive validation to ensure their reliability and effectiveness in diagnostics, underscoring the need for comprehensive testing across a broad range of conditions.

2.3 PCR and NGS methods

PCR and NGS-based methodologies have been instrumental in assessing DNA methylation status, each with distinct advantages for ctDNA biomarker analysis. PCR-based techniques, such as MSP,⁶⁷ methylation-sensitive high-resolution melting analysis,^{68, 69} MethyLight PCR,⁷⁰ one-step methylation-specific PCR (OS-MSP),⁷¹ methylation-specific PCR-coupled liquid bead array,⁷² real-time methylation-specific PCR (RT-MSP), and multiplex nested methylation-specific PCR,⁷³ are pivotal for quantifying ctDNA methylation. These methods, which utilize specific primers for methylated and nonmethylated DNA, are designed for separate PCR reactions targeting known sequences,⁷⁴ thereby enhancing sensitivity and specificity. ddPCR emerges as a cutting-edge technique, leveraging nanoscale water-in-oil droplet technology for the quantitative detection of DNA methylation (Figure 3B). This approach, notable for its high sensitivity, accuracy, and the ability to provide absolute quantification without a standard curve, is particularly suited for cancer liquid biopsy applications. ddPCR's simplicity, rapid setup, and minimal sample requirements, coupled with its remarkable sensitivity, offer significant advantages over NGS methods.⁷⁵ Furthermore, ddPCR's lack of need for complex bioinformatics analysis, alongside its superior sensitivity compared with real-time PCR (RT-PCR), makes it an invaluable tool for detecting low-abundance targets like methylated ctDNA with greater precision and accuracy.⁷⁶ Building on this foundation, researchers have developed “TriMeth,” a tumor-agnostic digital PCR assay targeting three CRC-specific methylation markers (C9orf50, CLIP4, KCNQ5), demonstrating high specificity and sensitivity in a cohort of patients undergoing surgery for colorectal liver metastases (CRLM).⁷⁷ These advancements underscore the evolving landscape of molecular diagnostics, where PCR and NGS-based methods, particularly ddPCR, play a crucial role in enhancing the detection and analysis of ctDNA methylation, paving the way for more precise, efficient, and personalized cancer diagnostics and monitoring.

NGS methods, including MethylCap-Seq,⁷⁸ MeDIP-Seq,⁷⁹ and MBD-seq,⁸⁰ are pivotal for detecting methylation levels across multiple genes in ctDNA. In particular, Xin et al.⁸¹ used NGS technology to analyze the methylation patterns of tumor suppressor genes in ctDNA from pancreatic cancer patients and found that it has a broad prospect as a noninvasive detection and monitoring method. While pyrosequencing offers precise examination of methylated regions, its high cost limits widespread application. Methylation-sensitive high resolution melting presents a cost-effective, rapid, and accurate PCR-based alternative. The MethyLight PCR technique, leveraging fluorescence-labeled probes with bisulfite-treated DNA, enables specific detection of DNA methylation patterns.²⁰ Despite the potential of NGS for predicting cancer recurrence through ctDNA analysis, the complexity and expense of these methods pose challenges for routine clinical application. Addressing this, Jin et al.⁸² introduced a cost-effective, single-tube methylation-specific quantitative PCR (mqMSP) method utilizing a panel of 10 methylation markers, capable of detecting as low as 0.05% tumor DNA in plasma. This approach enhances liquid biopsy's predictive accuracy for cancer recurrence, surpassing traditional methods and even the methylated Septin 9 (mSEPT9) assay in detecting early and precancerous polyps.⁸² The US FDA has approved Epi proColon and its updated version, Epi proColon 2.0, for CRC screening, utilizing DNA hypermethylation analysis.⁸³ These tests, based on RT-PCR for mSEPT9 DNA detection, offer higher diagnostic sensitivity than conventional markers and fecal occult blood testing.^83-85 Innovative techniques like “CancerDetector,”⁸⁶ which simulates collective methylation status across CpG sites, and GRAIL's prototype methylation technology, demonstrate the evolving landscape of cancer detection. These methods enable early-stage identification and precise tumor origin determination,⁸⁷ showcasing the potential of methylation analysis in enhancing cancer diagnostics. Furthermore, targeted and shallow genome sequencing, as developed by Nguyen et al.,⁸⁸ integrates multiple analytical dimensions, offering a comprehensive and cost-effective approach for early cancer detection and localization. This multimodal method, suitable for broad screening, underscores the shift toward more accessible and efficient diagnostic technologies. In summary, advancements in PCR and NGS-based methylation analysis are revolutionizing cancer detection, moving toward more sensitive, specific, and cost-effective methodologies. These developments hold promise for improving early diagnosis, guiding treatment decisions, and enhancing the precision of cancer treatment.

Current efforts to enhance the sensitivity of liquid biopsies for cancer are primarily focused on refining sequencing and in vitro analytical steps. However, the underlying challenge is the low abundance of ctDNA in blood samples. Martin-Alonso et al.⁸⁹ tackled this issue by developing a dual-component priming agent that combines lipid nanoparticles and monoclonal antibodies with cfDNA, effectively reducing macrophage phagocytosis and nuclease degradation of cfDNA. This innovation led to a tenfold increase in ctDNA yield and improved the sensitivity of tumor analysis using ctDNA from less than 10–75% by administering the agent intravenously 1−2 h before sample collection. The agent's effect on cfDNA is transient, similar to that of contrast agents, necessitating further research to confirm its efficacy and tolerability in humans. Recent advancements in ctDNA technology have concentrated on enhancing sequencing sensitivity, identifying variations in methylated sites among individuals, and discovering cancer-specific cfDNA methylation biomarkers. Analytical methods such as receiver operating characteristic curve analysis, t-tests, and the Kaplan–Meier method have been employed for disease-free survival (DFS) analysis, alongside Cox proportional hazards regression and CpG differential methylation analysis.^90-92 These efforts aim to improve the accuracy and potential of ctDNA methylation detection, as summarized in Table 1. The utilization of marker combinations offers greater accuracy than individual markers. However, many current methylation detection methods are either too complex for routine clinical application or offer lower sensitivity than alternative ctDNA detection methods. This underscores the need for a streamlined process that ensures high sensitivity and specificity while maintaining broad applicability. Liquid biopsy technology has been made possible by the advancement of microfluidics and nanotechnology, enabling highly sensitive detection of circulating biomarkers such as ctDNA and miRNA.^93-95 ctDNA methylation analysis and liquid biopsy technology have shown great potential for cancer diagnosis, prognosis monitoring, and treatment guidance. With the advancement of nanotechnology, microfabrication technology, and biosensing technology, these technologies will achieve higher sensitivity, specificity, and portability in the future, providing more effective tools for cancer management.^{4, 5, 96, 97}

The ctDNA methylation detection method has great potential in early cancer detection and recurrence monitoring, but its widespread application still faces multiple technical challenges. Specifically, these challenges include DNA fragmentation and loss, which are particularly evident in the key step of methylation analysis and may affect the accuracy of the results; sensitivity and specificity issues, especially in the early stages of cancer, where ctDNA concentrations are low, increasing the detection difficulty, and nontumor-derived cfDNA may also interfere with the detection; limitations in low methylation detection, some methods cannot accurately determine low methylation, limiting their application; tissue specificity, the differences in methylation patterns between different tumor types and tissues make it difficult to develop universal detection methods; background methylation noise, which may interfere with detection results; the lack of standardization in detection methods, there is currently no uniform standard to evaluate the performance of different methods; sample handling and storage issues, which have a significant impact on the results; cost and efficiency considerations, PCR-based methods are less expensive but have limited detection sites, while NGS-based methods are expensive and time-consuming; the need for personalized detection, which requires more resources and time for preliminary research and validation; and the challenge of rapid technological updates, clinical laboratories need to keep up and validate new technologies. In response to these challenges, researchers are constantly striving to improve and optimize ctDNA methylation detection methods. For example, they are developing more sensitive detection technologies, optimizing sample handling and storage conditions, standardizing operating procedures, and exploring personalized detection protocols to enhance the accuracy and reliability of detection. Additionally, with the advancement of big data and artificial intelligence technologies, it may be possible to automatically analyze and interpret methylation data using machine learning algorithms, further improving detection efficiency and accuracy.

3.1 Challenges and coping strategies

Although liquid biopsy technology shows great potential in cancer management, many challenges still need to be overcome before it can be widely used in the clinic. These challenges include quality control of sample collection, processing, and preservation, as well as standardization and refinement of detection techniques.⁴ The main challenges of ctDNA methylation testing as a biomarker in liquid tumor biopsies are summarized as follows. Sensitivity and specificity issues: ctDNA content in the blood is extremely low, especially in the early stages of the tumor, and the detection sensitivity is difficult to improve. At the same time, the contamination of normal cell DNA also affects the specificity. Differences in sample collection and processing: For different types and different periods of tumors, there is no unified standard process for sample collection, ctDNA extraction and amplification, which affects the consistency and repeatability of detection results. Tumor heterogeneity: There is heterogeneity between tumor tissue and ctDNA, and between primary tumor and metastatic tumor, resulting in inconsistent detection results. Technical cost and accessibility limitations: The high cost of highly sensitive methylation sequencing technology limits its widespread clinical utilization. In view of the above challenges, researchers are actively exploring and implementing a series of solutions. In terms of sensitivity and specificity, the introduction of more advanced sequencing techniques and noise reduction methods is expected to significantly improve the accuracy and reliability of ctDNA methylation detection. In terms of sample processing and standardization, the development of a unified sample collection and processing protocol, combined with automated instruments and equipment, will help improve the stability and repeatability of test results. Aiming at the challenge of tumor heterogeneity, the application of multiomics combined analysis and big data technology will provide strong support for more comprehensive analysis of tumor heterogeneity. Finally, in terms of technology cost and accessibility, by optimizing sequencing technology and data processing processes, and developing portable, rapid detection equipment, it is expected to reduce detection costs and improve the popularity and accessibility of technology.

Sensitivity and specificity are the most critical and challenging problems in ctDNA methylation detection. Solving this problem requires not only highly sensitive sequencing techniques, but also a combination of advanced bioinformatics analysis and machine learning algorithms to distinguish between tumor-specific methylation signals and background noise. The use of sequencing technologies such as WGBS and RRBS, combined with advanced enrichment and noise reduction methods such as cfDNA–RBS technology, can significantly improve the detection sensitivity and specificity. At the same time, the development of a highly sensitive targeted sequencing platform, combined with multimethylation PCR and noise reduction technology, and the construction of a scoring model based on machine learning algorithms, can accurately classify and evaluate methylation signals, and further improve the accuracy of detection results.

3.2 Improvements in traditional methylation detection techniques

Methylation detection technologies, pivotal in understanding epigenetic modifications, can be categorized based on genome coverage into three main types: WGBS, RRBS, and methylation capture technology. WGBS offers comprehensive coverage of all methylation sites and, despite its high cost, is unparalleled in detecting low-frequency methylation signals when combined with targeted approaches. RRBS, while being the most cost-effective among NGS technologies, suffers from less stable sequencing data due to its reliance on MspI cleavage, covering approximately 10% of methylation sites. Methylation capture technology, positioned cost-wise between WGBS and RRBS, employs hybridization probes to enrich methylation-specific fragments, resulting in more stable data than RRBS. WGBS stands out as the optimal screening method, capable of assessing all CpG sites. We have advanced the ctDNA–WGBS technique, proving its efficacy in analyzing comprehensive methylation patterns across the entire genome with minimal ctDNA quantities.⁵⁰ Remarkably, this method requires only 200 µL of plasma, significantly less than the traditional 5–20 mL. As detailed in Figure 3C, we streamlined the library preparation process by integrating end repair, dA tailing, adapter ligation, and bisulfite conversion into a single tube, enhancing efficiency. Additionally, we replaced agarose gel capture with bead-based methods to improve the recovery ratio. This refined approach allows for library preparation from as little as 1 ng of input, ensures unbiased genome-wide coverage, and incorporates advanced computational strategies to filter out noise from low recurrent fragments and nontumor-derived ctDNA. Consequently, it facilitates highly specific and sensitive detection of early-stage BC across various patient cohorts and distinguishes between different cancer molecular subtypes. Despite its widespread use, bisulfite sequencing, the gold standard for DNA methylation detection, faces significant challenges, particularly with limited samples like cfDNA in blood. Bisulfite treatment can degrade up to 84−96% of DNA, complicating the analysis of scarce samples. Moreover, bisulfite sequencing indirectly detects 5mC and 5hmC, converting unmodified cytosines to uracil, which diminishes sequence complexity, degrades sequencing quality, and hampers efficient targeting.

3.3 Improvement of bisulfite-free technology

In response to the limitations of traditional sequencing methods, several bisulfite-free technologies have been developed. Shen et al.¹⁰³ introduced cfMeDIP-seq, a novel approach that combines immunoprecipitation with high-throughput sequencing to specifically enrich methylated DNA fragments in cfDNA, thereby enhancing detection efficiency. Similarly, researchers at the University of Pennsylvania have developed APOBEC-coupled epigenetic sequencing. This technique replaces the chemical transformation of bisulfite with an enzymatic process, using the APOBEC deaminase enzyme to effectively distinguish between cytosine states without causing DNA damage.¹⁰⁴ Song et al.¹⁰⁵ discovered a sulfite-free, base-resolution sequencing method known as TET (ten-eleven translocation)-assisted pyridine borane sequencing. This method employs the TET enzyme to convert 5mC and 5hmC into a third modification, which is then chemically converted to thymine for direct sequencing. Yuming Lu's team has developed a methodology for the direct detection of 5mC using single molecule real-time (SMRT) sequencing, providing a comprehensive examination of the kinetic signals of DNA polymerase and the sequence context of each nucleotide. This approach achieved a sensitivity and specificity of 90 and 94%, respectively, with a 99% correlation to overall methylation levels compared with sulfite sequencing.¹⁰⁶ Nanopore sequencing offers direct detection of methylated cytosines without the need for bisulfite conversion. Adaptive sampling, a method that enriches regions of interest by removing off-target regions during sequencing, has been utilized by Danny et al.¹⁰⁷ in conjunction with the Oxford Nanopore platform for targeted long-read sequencing. This approach allows for the identification of pathogenic substitutions, structural variants, and methylation differences using a single data source. However, the requirement for unamplified genomic DNA may limit the application of single-molecule approaches like SMRT-seq and Nanopore sequencing in rare clinical samples. To address this, Li et al.¹⁰⁸ developed NT-seq, a method that sequences multiple DNA methylations based on chemical transformation, enabling the detection of various DNA epimethylations at the genome-wide level. Most recently, Wang et al.¹⁰⁹ introduced direct methylation sequencing (DM-Seq), an all-enzymatic method that uses nanogram quantities of DNA for single-base resolution analysis of 5mC. DM-Seq offers a nondestructive, reliable, and direct detection method for 5mC, overcoming the traditional challenges of distinguishing between 5mC and 5hmC.

3.4 Optimization algorithms and models

GRAIL's cfDNA-targeted methylation liquid biopsy assay represents a significant advancement in cancer detection, capable of identifying and localizing over 50 types of cancer, including those with high mortality rates and lacking screening guidelines. This assay achieves a remarkable specificity of 99.3% and can accurately identify the tissue source of the cancer signal with a 93% accuracy rate when a cancer signal is detected.¹¹⁰ However, the predictive values calculated by GRAIL currently rely on incidence rates rather than prevalence rates, which may not be the most logical approach. Fiala and Diamandis¹¹¹ have critiqued the company's study for comparing asymptomatic subjects with known cancer patients, arguing that this inflates the test's positive predictive value and does not accurately reflect its use in screening the general population. To address the high costs associated with high-throughput sequencing, Li et al.¹¹² proposed DISMIR, a novel method incorporating a “switching region” to identify cancer-specific differentially methylated regions. DISMIR, leveraging deep learning models, has shown high precision and robustness in detecting HCC from plasma cfDNA WGBS data, even at ultra-low sequencing depths. Its insensitivity to individual CpG site methylation status changes suggests resistance to WGBS technical noise, positioning DISMIR as a promising tool for early cancer detection.

In the Circulating Cell-free Genome Atlas Substudy,¹¹³ researchers evaluated 10 sequencing analysis technologies targeting cfDNA for early tumor screening. Findings indicated that cfDNA methylation is the most effective marker for predicting cancer signal sources, with genome-wide methylation classifiers exhibiting the highest sensitivity. This approach has demonstrated the capability to detect common cancer signals across more than 50 cancer types with 99.5% specificity and accurately predict cancer origins. Wang and colleagues¹¹⁴ introduced mutation capsule plus (MCP) technology, enhancing liquid biopsy assays for HCC detection compared with current assays or evaluating methylation or mutation markers alone. MCP technology holds promise for discovering and validating noninvasive cancer detection biomarkers.

Chemi et al.³⁰ developed a methodology for genome-wide DNA methylation analysis applicable to patient-derived models and cfDNA, improving SCLC detection sensitivity and distinguishing between different progression stages. David et al.’s¹¹⁵ “HCC-detect” classifier, based on four CpG sites, identified 98% of HCC samples with 100% specificity, differentiating HCC from various liver conditions. Their epiLiver assay, utilizing a focused multiplexed high-throughput next-generation sulfite sequencing approach, classified HCC patients with 84.5% sensitivity and 95% specificity. Building on a comprehensive tissue methylation atlas from 521 noncancerous tissue samples across 29 major human tissues, Li et al.¹¹⁶ developed cfSort, a novel deep-learning model for precise tissue identification within cfDNA. cfSort outperformed existing methods in benchmark data, enhancing tissue deconvolution performance in cfDNA for disease detection and therapy monitoring. However, accurately quantifying tissue-derived cfDNA remains a challenge due to limited knowledge on tissue methylation and reliance on unsupervised techniques, highlighting the need for further research in this area.

In summary, as an important biomarker in tumor liquid biopsy, ctDNA methylation detection faces many challenges, but through continuous technological innovation and optimization, it is expected to become an important tool for early tumor screening and diagnosis in the future.

4.1 Early screening and diagnosis

In recent years, the exploration of ctDNA as a diagnostic tool for various cancers has garnered significant research interest, as outlined in Table 2. The application of ctDNA methylation is emerging as a promising strategy for the early detection of cancer. Cancer-specific methylation patterns, which appear at the early stages of tumor development and exhibit remarkable stability, provide a highly accurate signal for precise analysis.¹²² Klein et al.¹²³ reported a large-scale clinical validation of a methylation-based classification in 4077 participants, achieving a specificity rate of 99.5% for cancer detection. This underscores the potential of methylation-based ctDNA analysis to enhance early-stage cancer screening across various types. Research focused on early CRC detection through ctDNA methylation, particularly SEPT9 hypermethylation, has demonstrated high sensitivity and specificity.¹²⁴ This approach has shown promise in surpassing traditional screening methods, such as endoscopy, especially in high-risk groups for CRC.¹²⁵ The diagnostic efficacy of CRC has been significantly enhanced by combining methylation markers with current markers like carcinoembryonic antigen (CEA) and fecal immunochemical test, achieving higher sensitivity and improved AUC values.¹²⁶ Furthermore, a study by Brenne et al.¹²⁷ indicates that known methylation markers of CRC detected by liquid biopsy can be detected up to 2 years before clinical diagnosis, suggesting its potential for early screening of CRC. Yu et al.¹²⁸ developed a highly sensitive screening method for esophageal cancer using five candidate markers with significant hypermethylation, suggesting DNA methylation-based molecular detection as a novel approach for esophageal adenocarcinoma screening. Similarly, ctDNA methylation detection has shown high sensitivity in identifying early-stage LC, indicating its potential as a method for early diagnosis and screening.^{64, 129} Liang et al.’s¹³⁰ model, PulmoSeek, which incorporates 100 methylation features, demonstrated high sensitivity in detecting early-stage LC, outperforming existing diagnostic models and positron emission tomography-computed tomography in precision. In addition, Bu et al.¹³¹ evaluated the application of Septin9 gene methylation in the diagnosis of cervical cancer and prediction of pelvic lymph node metastasis, demonstrating its potential as a novel biomarker. Lung EpiCheck, a novel blood test based on six methylation markers, has shown high detection rates for early non-small cell lung cancer (NSCLC) and SCLC in high-risk populations.¹³² Additionally, the detection of NSCLC-specific ctDNA in urine through methylation analysis of CDO1 and SOX17 gene markers has been explored, indicating the feasibility of this noninvasive method for NSCLC detection.¹³³ In HCC research, ctDNA detection precedes imaging abnormalities and elevated alpha-fetoprotein (AFP) levels,¹³⁴ with hypomethylation near hepatitis B virus integration sites serving as an effective early screening method.¹³⁵ The hypermethylation status of additional genes, such as the CTCFL promoter and UBE2Q1 in ctDNA, is associated with HCC diagnosis and patient monitoring.^{136, 137} The detection of ctDNA methylation levels in serum/plasma samples from OC patients aids in early detection and guides individualized treatment.^{69, 138, 139} Promoter region methylation inactivating tumor suppressor genes is a significant factor in OC development,^{138, 140} warranting further investigation. Li et al.⁴² analyzed DNA methylation kinetics in consecutive CSF samples, demonstrating the high sensitivity and clinical potential of ctDNA methylation labeling for monitoring MB illness status. In conclusion, ctDNA methylation marker detection offers a noninvasive and effective means to improve adherence and participation in cancer screening, highlighting its significant potential in precision oncology.

4.2 Monitoring for recurrence and MRD

Recent advancements have highlighted the significant role of ctDNA methylation in cancer surveillance, particularly in detecting MRD and tumor recurrence. Michail et al.¹⁵¹ first introduced the concept of “ctDNA recurrence,” underscoring ctDNA's critical role in monitoring cancer. Postoperative or after curative therapy, the presence of MRD is a primary cause of tumor recurrence. Numerous studies have demonstrated the utility of ctDNA methylation markers in monitoring tumor recurrence and MRD (Table 3), offering advantages such as rapid turnaround time, low initial cost, and applicability in cases where tumor tissue is unavailable or inaccessible.³⁸ For instance, the detection of methylated BCAT1 and IKZF1 in blood has shown potential in identifying residual disease posttreatment, suggesting a need for treatment modification in advanced-stage illness.^{47, 152, 153} This is further supported by Wang et al.,²⁵ who developed a 5-marker ctDNA methylation model with 87.5% sensitivity and 94.12% specificity for predicting tumor recurrence in CRC patients. This model allows for earlier CRC recurrence detection than imaging techniques and serum CEA measurement, providing a basis for early treatment adjustment. Pedersen et al.¹⁵⁴ established a reference level for plasma BCAT1 and IKZF1 methylation markers, enhancing specificity and clinical accuracy for recurrence detection. Al Naji et al.¹⁵⁵ evaluated the predictive value of BCAT1/IKZF1 ctDNA methylation for CRC recurrence and found that its combination with clinical prognostic factors significantly improves prediction accuracy. Similarly, elevated levels of methylated SEPT9 ctDNA have been identified as a significant biomarker for CRC recurrence,⁶ offering superior potential over CEA.¹⁵⁶ Jin et al.’s⁸² mqMSP assay targeting the SEPT9 gene predicted CRC recurrence within 2 weeks of surgery, outperforming traditional CEA markers. A prospective clinical trial measuring ctDNA methylation levels of WIF1 and NPY pre- and postsurgery in stage II–III CRC patients found a higher 2-year DFS rate in ctDNA-negative patients,¹⁵⁷ indicating the potential of ctDNA methylation analysis in predicting postoperative MRD. Slater et al.¹⁵⁸ conducted a study and proposed that longitudinal sampling of MRD assessment can predict recurrence in stage I–III CRC, providing a basis for individualized treatment. Additionally, hypermethylation in ctDNA of genes like RASSF1A, COX2, and APC has been linked to increased tumor recurrence and poorer survival in HCC patients.¹⁵⁹ Research on OC patients has shown that DNA methylation in genes such as SFRP1, SFRP2, SOX1, and LMX1A is associated with a higher risk of recurrence and poorer overall survival (OS) outcomes.¹⁶⁰ Furthermore, ctDNA methylation detection offers promising avenues for identifying MRD and monitoring tumor progression. Parikh et al.¹⁸ provided an analytical method using a focused gene panel and methylation markers to assess MRD, eliminating the need for resected tumor tissue. This MRD assay, utilizing plasma alone, demonstrated high specificity and sensitivity in detecting recurrence, with further enhancements achieved by incorporating epigenomic markers and analyzing serial longitudinal samples.¹⁸ The ongoing NCT04068103 trial employs this assay to inform treatment decisions in stage IIA colon cancer,³⁸ showcasing the potential of a tumor-blind methodology in ctDNA-based MRD detection. These advancements highlight the dynamic monitoring of tumors using ctDNA methylation as an effective strategy for early detection of tumor recurrence and facilitating clinical decision-making.

4.3 Predicting treatment response and prognosis

The determination of prognosis is significantly influenced by ctDNA methylation models, which utilize blood DNA methylation to develop prognostic models. This approach enables clinicians to quickly identify high-risk individuals, providing timely and effective interventions for those at risk and offering less-intensive treatments or follow-up for those not at high risk. By doing so, it avoids the widespread use of traumatic therapies that can negatively affect patients and waste medical resources. Extensive research on ctDNA's ability to predict cancer prognosis (Table 4), including in OC, CRC, and prostate cancer, has yielded improved predictive outcomes.^{8, 163-165} Studies have shown a correlation between ctDNA dynamics and treatment response,^{147, 166} with Boeckx et al.¹⁶⁷ demonstrating that changes in ctDNA concentrations can reflect individual treatment responses by combining mutated and methylated ctDNA biomarkers. For instance, a study found significant reductions in ctDNA methylation levels of HPP1 after induction chemotherapy,¹⁶⁸ suggesting its potential as a prognostic marker. The challenge of preventing metastatic spread in locally advanced rectal cancer highlights the need for effective prognostic factors for pretreatment metastatic progression. In this context, ctDNA quantification by assessing BCAT1 and IKZF1 methylation levels has shown that posttreatment ctDNA detection is associated with disease progression,¹⁶⁹ offering valuable insights for clinical decision-making regarding further treatment or monitoring. Postoperative increases in ctDNA methylation levels are associated with poor tumor prognosis. For example, SEPT9 methylation levels in CRC patients dropped significantly after surgery,¹⁷⁰ while postoperative detection of plasma SEPT9 hypermethylation was linked to higher mortality rates.¹⁷¹ Similarly, methylation status has been identified as a valuable biomarker for predicting NSCLC prognosis,^{172, 173} with studies showing that specific methylation patterns are associated with unfavorable outcomes.^{174, 175} In high-grade serous ovarian cancer (HGSOC), the prognostic significance of methylation in genes such as SOX17,¹⁷⁶ CST6,¹⁷⁷ BRMS1,¹⁷⁸ RASSFIA,⁶⁹ and ESR1¹⁷⁹ in plasma ctDNA has been demonstrated. Additionally, GSTP1 promoter methylation in ctDNA has been analyzed^{180, 181} and has been proven useful for prognosticating metastatic castration-resistant prostate cancer.¹⁸² Tian et al.²⁶ developed a cmp score based on methylated ctDNA composition to predict ENKTL prognosis, highlighting the need for further clinical trials to validate its practical application. In rhabdomyosarcoma, RASSF1A-M methylation in cfDNA was significantly associated with poor prognosis.¹⁸³ While the prognostic value of ctDNA methylation patterns has been observed across various cancers, the refinement and validation of methylation markers for prognostic assessment continue to be crucial areas of research.

4.4 Assessing tumor burden

The short half-life of cfDNA directly correlates with tumor cell status, offering a unique opportunity for continuous monitoring of cancer burden during treatment. Recent studies have explored the utilization of ctDNA methylation markers for assessing tumor burden (Table 4). Melton et al.¹⁹¹ proposed a new statistical method that quantifies the cancer-specific methylation patterns in cfDNA to estimate ctDNA abundance, which helps noninvasive assessment of tumor burden. Bjerre et al.¹⁰¹ detected a positive correlation between tumor burden and plasma methylated ctDNA levels, suggesting that methylated ctDNA could serve as a biomarker to identify patients with mPCa with high tumor volume, potentially guiding therapy decisions for patients at this stage.¹⁰¹ However, these findings are limited by the small number of mPCa patients in the study, necessitating further validation in a larger cohort. In another study, methylation mDETECT levels were shown to predict elevated prostate-specific antigen (PSA) levels, indicating tumor burden in PSA-negative tumors.¹⁸² In addition, a study by Angeli-Pahim et al.¹⁹² indicates that quantifying methylation of ctDNA can effectively monitor changes in tumor burden in HCC patients without the need for tumor biopsy. Tran et al.¹⁹³ analyzed 366 plasma ctDNA samples from 85 patients, using CT-based three-dimensional annotated tumor volumes to identify early-stage NSCLC patients at risk for recurrence. This integrated imaging analysis could stratify risk groups for clinical outcomes. Symonds et al.’s¹⁶⁹ research quantified ctDNA by assessing BCAT1 and IKZF1 methylation levels, demonstrating the potential of this approach to evaluate tumor burden and treatment response. In a study of 175 patients with pretreatment-positive CRC ctDNA, levels increased with disease progression and were significantly correlated with tumor diameter and volume. Methylation levels of BCAT1 and IKZF1 DNA decreased substantially in 48 fully treated cases, with undetectable levels in the majority (87.5%) of patients. This analysis revealed a consistent trend of increasing ctDNA levels with tumor mass.¹⁶⁹ To investigate the potential of using blood mSEPT9 to evaluate treatment response in CRC, Song et al.¹⁷¹ found that mSEPT9 levels were associated with tumor size, unlike CEA levels. This suggests that ctDNA methylation markers offer a promising approach for real-time monitoring of tumor dynamics and assessing the effectiveness of treatment strategies.

4.5 Determine the source of the organization

Epigenetic biomarkers, characterized by their tissue-specific DNA methylation patterns, have emerged as a promising method for pinpointing the origin of ccfDNA. Despite their potential, further targeted research is essential to deepen our understanding. Unlike tissue biopsies, which risk DNA integrity during preparation, ctDNA samples from plasma/serum remain intact,¹⁹⁴ allowing for the identification of the tissue of origin through methylation profile differences.^{195, 196} Studies like CancerSEEK and Methylation Group GRAIL have underscored ctDNA's capability in identifying the tissue origin in disseminated malignancies without a clear primary focus, achieving up to 83% accuracy.¹⁹⁷ “Plasma DNA tissue mapping” through genome-wide analysis supports the feasibility of determining cfDNA tissue origin by analyzing methylation patterns across tissues.¹⁹⁸ Additionally, research using tissue-specific DNA methylation biomarkers and digital PCR to measure DNA concentrations in plasma has shown clinical applications in metastatic CRC, particularly for hepatic and colonic origins.¹⁹⁹ In a study of 26 HCC patients, the association of 5hmC modifications in a select list of genes provided estimates for the origin of these modifications, indicating that 5hmC in cfDNA correlates with tissue origin.²⁰⁰ Tian et al.’s²⁶ research revealed a strong correlation between differential methylation marks in ENKTL plasma and tissue DNA methylation, suggesting ctDNA methylation patterns could indicate ENKTL tissue methylation status. Given the unique methylation patterns across tissues, ctDNA methylation analysis holds promise for identifying specific tissue sources, offering new avenues for detecting malignancies where traditional biopsy methods are inadequate.

4.6 Predictive immunotherapy

Wen et al.²⁰¹ found that the combination of natural killer cell activity and methylation status of HOXA9 ctDNA in NSCLC patients receiving programmed cell death protein 1 (PD-1)/its ligand PD-L1 inhibitor therapy has prognostic value. The treatment of diffuse large B-cell lymphoma (DLBCL) heavily relies on intensive anthracycline-based chemoimmunotherapy. Tumor cells treated with azacitidine have shown increased sensitivity to chemoimmunotherapy.²⁰² Brem et al.²⁰³ have initiated a randomized trial incorporating oral hypomethylating agents for older DLBCL patients, aiming to reduce injection frequency. This trial also explores methylation patterns in normal T cells and ctDNA to understand the effects of demethylating drugs on both tumor DNA and healthy cells.²⁰³ In NSCLC, systemic therapy often leads to rapid responses but is marred by drug resistance and vulnerability to distant metastases. The analysis of ctDNA has facilitated the identification of drug resistance linked to CYP2D6 deficiency and the evaluation of treatment responses.²⁰⁴ DNA methylation deficiency, associated with chromosomal instability, may predict poor responses to immunotherapy.²⁰⁵ Further analysis revealed a positive correlation between CYP2D6 methylation and lymphocyte infiltration in NSCLC samples, suggesting that higher CYP2D6 methylation levels indicate increased immunogenicity. This insight led to the identification of MEK inhibitors as potential treatments for CYP2D6 deficiency.²⁰⁴ These findings highlight innovative approaches to overcoming resistance in systemic NSCLC therapy and underscore the importance of advancing new therapeutic strategies. While predicting neoantigens directly remains challenging, ctDNA methylation offers an indirect, probabilistic indicator of immunogenicity, paving the way for personalized cancer treatment.

4.7 Clinical trials employing ctDNA methylation detection

ctDNA methylation detection has emerged as a promising tool for early cancer diagnosis, monitoring, and prognosis. This summary focuses on ongoing and completed clinical trials that utilize ctDNA methylation detection across various cancer types, summarized in Table 5 by pathway and mechanism of action. Trials NCT05801263²⁰⁶ and NCT05801276²⁰⁷ investigate the diagnostic sensitivity and specificity of CDO1 and HOXA9 methylation assays. These studies aim to validate the clinical application of these biomarkers for early detection and prognosis in OC. Additionally, NCT05578625²⁰⁸ focuses on monitoring MRD in multiple myeloma (MM) patients using plasma ctDNA methylation sequencing. The primary outcome is the completion of sample collection, with the goal of tracking clonal evolution patterns during disease progression. Furthermore, several trials (NCT05536089,²⁰⁹ NCT05558436,²¹⁰ NCT03737591,²¹¹ NCT06347887,²¹² NCT03828396,²¹³ NCT03737539,²¹⁴ NCT05508503,²¹⁵ NCT05904665,²¹⁶ NCT05954078²¹⁷) evaluate the efficacy of ctDNA methylation assays for monitoring postoperative relapse, assessing adjuvant chemotherapy efficacy, and early detection of CRC and adenomas. These studies focus on improving diagnostic accuracy, DFS, and ctDNA clearance rates. Similarly, trials NCT03685669,²¹⁸ NCT03651986,²¹⁹ NCT06358222,²²⁰ NCT04253509,²²¹ NCT03181490,²²² and NCT03634826²²³ aim to develop noninvasive methods for detecting lung cancer-specific ctDNA in blood. These studies focus on improving the specificity of early lung cancer screening, differentiating benign and malignant pulmonary nodules, and monitoring postoperative recurrence. Moreover, NCT03483922²²⁴ aims to develop noninvasive biomarkers based on ctDNA methylation changes for HCC detection. The primary outcome is the calculation of M scores and HCC probability scores. Additionally, NCT05858242²²⁵ investigates the application of ctDNA methylation in early screening, diagnosis, and postoperative monitoring of BC. The study aims to establish a detection system and evaluate the predictive effect of ctDNA methylation on postoperative prognosis. Last, trials NCT03922230²²⁶ and NCT04511559²²⁷ focus on detecting hydroxymethylation and ctDNA methylation in esophageal squamous cell carcinoma (ESCC) and gastric cancer, respectively. These studies aim to identify methylation markers associated with early cancer detection and prognosis, with the collective goal of enhancing early cancer detection, improving diagnostic accuracy, and providing better monitoring of cancer recurrence through innovative ctDNA methylation detection methods. The integration of ctDNA methylation analysis with other diagnostic tools holds significant potential for transforming cancer diagnostics and patient management.

These studies highlight the pivotal role of ctDNA methylation and liquid biopsy in early cancer detection, prognosis evaluation, and disease monitoring. The advancement of these technologies offers novel tools and approaches for precise diagnosis and personalized medicine in cancer care.

Although ctDNA methylation detection holds promise, several challenges must be addressed before its widespread adoption in clinical practice. Key challenges include developing reproducible assays with high sensitivity and specificity and validating biomarker efficacy in prospectively collected target populations. The determination of the optimal panel for various stages of cancer diagnosis, recurrence monitoring, treatment response prediction, and prognosis, given the proven utility of specific promoter regions, remains unresolved. Identifying methylation markers that combine satisfactory sensitivity and specificity with the capability for rapid detection in large sample sizes is essential. The development of prognostic methylation markers faces numerous obstacles, partly due to difficulties in standardizing detection techniques across studies and identifying universally accepted prognostic methylation sites. Moreover, current detection methods often rely on sulfite conversion, which may damage ctDNA.^{230, 231} It is crucial to acknowledge that DNA methylation patterns vary across different ethnicities and environmental exposures before applying ctDNA methylation assays in clinical settings. The presence of ctDNA methylation in a significant number of healthy controls,^{232, 233} along with the potential for false positives or preoperative false negatives,⁸² necessitates further optimization. Effective screening methods must achieve high specificity to minimize the risk of overdiagnosis,²³⁴ especially given the low incidence of cancer. The manual nature of current assays, including sample preparation, DNA extraction, and bisulfite conversion, introduces the risk of errors and inefficiencies. The reliance on retrospective data and the lack of serial plasma samples in several studies^{25, 66, 147, 235} call for validation in larger, more homogeneous cohorts with consistent follow-up to ensure the robustness, reproducibility, and reliability of reported results. The high cost of sequencing per sample poses another barrier to cost-effective screening, highlighting the need for simpler, more affordable methods for identifying and measuring methylation levels. Observational studies often suffer from variability in treatment and follow-up protocols, limiting their applicability compared with structured prospective clinical trials. Ultimately, the full clinical application of ctDNA methylation detection depends on demonstrating its clinical validity and utility, paving the way for its significant impact on genomic oncology and the management of cancer patients.