From concept to clinic: emerging paradigms and novel insights into liquid biopsy

Foundation and clinical motivations

Liquid biopsy is based on the premise that circulating molecular biomarkers in disease states systematically diverge from those in healthy individuals, and therefore can reflect perturbations in genetic, epigenetic, and regulatory networks. These alterations are both disease-specific and stage-dependent, and they mirror the dynamic molecular evolution that accompanies disease initiation, progression, and therapeutic response. Seminal observations of fetal DNA in maternal circulation provided early evidence that pathological tissues, including tumors and virally infected or degenerating cells, release nucleic acids into the bloodstream, thereby enabling indirect molecular interrogation of otherwise inaccessible sites⁴. These circulating signals encompass diverse molecular phenotypes, including somatic mutations, chromosomal rearrangements, aberrant methylation patterns, and viral integration events.

The clinical utility of this paradigm is exemplified by non-invasive prenatal testing (NIPT), in which analysis of cell-free fetal DNA has transformed prenatal screening for chromosomal aneuploidies and substantially decreased the need for invasive diagnostic procedures. Similarly, in infectious diseases such as hepatitis B, hepatitis C, and AIDS, circulating viral nucleic acids are routinely used to guide antiviral therapy, assess treatment efficacy, and monitor adherence⁵. Together, these circulating biomarkers constitute a molecular “fingerprint” providing near-real-time insight into ongoing pathological processes, thus underscoring the central role of liquid biopsy in precision medicine.

In oncology, early detection is a primary clinical goal. The ability to detect trace amounts of ctDNA in asymptomatic or early-stage patients has promise for diagnosing malignancies in curable stages. Beyond diagnosis, longitudinal ctDNA profiling enables real-time monitoring of clonal dynamics, early identification of resistance mechanisms, and informed selection of subsequent therapeutic strategies⁶. Importantly, liquid biopsy enables repeated sampling with minimal patient burden, thereby overcoming the limitations of serial tissue biopsies, which are often invasive and impractical. Emerging fragmentomic approaches that exploit tumor-specific differences in cfDNA fragment size and genomic breakpoints further enhance analytical sensitivity and biological resolution⁶.

A central ongoing debate concerns biomarker strategy, i.e., whether to prioritize highly specific, disease-defined markers or broader, pan-disease signatures. Canonical examples such as EGFR mutations in non-small cell lung cancer or BCR-ABL fusions in chronic myeloid leukemia provide precise diagnostic and therapeutic value⁷. In contrast, pan-cancer approaches, such as assays targeting shared methylation or fragmentomic patterns, might be better suited for population-scale screening when the tissue of origin is unknown⁸. To successfully integrate these complementary strategies, careful optimization of sensitivity, specificity, and cost-effectiveness will be necessary, thus underscoring the need for rational assay design aligned with clinical intent.

Heterogeneity in disease applications

Metastatic malignancies, including advanced lung, pancreatic, and breast cancers, comprise diverse subclonal populations with distinct genetic and epigenetic landscapes⁹. Single-site tissue biopsies, which capture only a limited spatial snapshot, can therefore overlook clinically relevant subclones bearing resistance-associated alterations. In contrast, longitudinal profiling of ctDNA enables systemic and dynamic assessment of tumor evolution, and can reveal adaptive responses to therapeutic pressure in near real time. This comprehensive view supports more informed treatment stratification, facilitates early detection of emerging resistance, and provides a rationale for timely therapeutic switching or combination strategies aimed at delaying disease progression.

Importantly, the utility of liquid biopsy is not confined to conditions with extensive genomic heterogeneity. In disorders with more constrained genetic variability, circulating biomarkers can still yield clinically actionable insights. For example, pre-eclampsia arises from complex maternal-fetal interactions rather than extensive clonal diversification. Nevertheless, placental-derived cfDNA can exhibit early, disease-associated epigenetic signatures or fragmentomic patterns that signal elevated risk, thus enabling earlier clinical intervention¹⁰. Similarly, neurodegenerative diseases such as Alzheimer’s disease and Parkinson’s disease lack the overt clonal dynamics typical of cancer, thus complicating biomarker discovery. However, emerging evidence suggests that subtle yet reproducible cfDNA alterations may precede overt clinical manifestations, thereby offering a potential avenue for presymptomatic detection and disease monitoring. These findings exemplify how the adaptability of liquid biopsy supports its transformative potential in neurodegenerative research, by providing novel biomarkers that might reshape diagnostic pathways.

ctDNA vs. total cfDNA: complementary strategies

Liquid biopsy research in oncology has traditionally centered on ctDNA, with a goal of identifying tumor-specific genetic or epigenetic alterations. This targeted paradigm enables precise detection of clinically actionable mutations, such as EGFR variants in non-small cell lung cancer or ALK/ROS1 rearrangements, thereby directly informing personalized therapeutic decisions¹¹. However, in early-stage disease, ctDNA often constitutes only a minute fraction of total circulating cfDNA and therefore poses substantial analytical challenges. To address this limitation, ultrasensitive platforms, including digital droplet PCR and error-corrected targeted next-generation sequencing, have been developed to enhance detection sensitivity and specificity.

In parallel, an emerging paradigm emphasizes analysis of total cfDNA rather than a focus on only tumor-derived fragments. This holistic approach interrogates global features such as fragmentation profiles, DNA methylation landscapes, and chromatin accessibility patterns across the entire cfDNA pool. By capturing system-wide molecular perturbations, total cfDNA analysis can reveal early disease-associated signals, infer tissue of origin, and reflect broader pathological states, including inflammation or multi-organ involvement. Although this strategy might sacrifice mutation-level specificity, it offers a more integrative view of disease biology. As analytical technologies continue to mature, the key challenge is no longer choosing between ctDNA-focused and global cfDNA approaches, but developing integrative frameworks that harness their complementary strengths to deliver more sensitive, informative, and clinically actionable liquid biopsy assays.

Immunogenomic insights and microbiome profiling

Beyond tumor- or tissue-derived biomarkers, the immunogenomic information provided by circulating cfDNA provides an additional dimension for understanding disease progression and therapeutic response. Profiling circulating T-cell receptor and B-cell receptor repertoires provides insights into adaptive immune dynamics in cancer, infection, and immune-mediated disorders¹². Clonal expansion or contraction of specific lymphocyte populations detectable in plasma has been associated with responses to immunotherapies, including immune checkpoint blockade¹³. Beyond serving as descriptive immune snapshots, repertoire features can be mechanistically informative. Antigen-driven activation may lead to clonal expansion and diminished diversity, whereas effective therapy can alter these dynamics over time. In contrast, loss of dominant clones may indicate immune escape. In agreement with this possibility, recent data have suggested that circulating T-cell receptor signals can track tumor immune dynamics during checkpoint blockade and complement tumor-derived cfDNA in monitoring treatment response¹⁴. Integrating immune repertoire features with tumor-derived cfDNA may therefore enable more precise monitoring of immune activation, exhaustion, and treatment responsiveness, and consequently inform personalized immunotherapeutic strategies.

In parallel, circulating microbial cfDNA has emerged as a complementary and increasingly informative analyte. Disruptions in microbial homeostasis across the gut, oral cavity, or other niches can allow bacterial or viral DNA fragments to enter the bloodstream, thereby reflecting systemic dysbiosis or occult infection. Accumulating evidence suggests that specific microbial signatures influence tumor biology and modulate responses to immunotherapy by shaping the tumor microenvironment and host immune tone¹⁵. Consequently, longitudinal profiling of microbial cfDNA may aid in predicting therapeutic efficacy, detecting treatment-related complications, and contextualizing inflammatory states.

Novel frontiers: cf-mtDNA fragmentomics in early tumor screening

Because of the high copy number of mitochondrial genomes per cell, cf-mtDNA is often detectable at higher baseline levels than nuclear cfDNA, even in early-stage malignancies. Beyond copy number changes, recent studies have demonstrated that cf-mtDNA fragmentomic features, including fragment length distributions, 5′ end motifs, and sequence preferences, can robustly distinguish tumor-associated signals from healthy backgrounds¹⁶.

Multicenter studies have shown that cf-mtDNA fragmentomics enables sensitive detection of early-stage colorectal cancer and advanced adenomas, and frequently outperforms conventional DNA methylation-based assays¹⁷. Early neoplastic lesions are consistently characterized by features such as relatively short fragment sizes, altered 5′ end base composition, and diminished motif diversity. Importantly, integrating cf-mtDNA fragmentomics with established nuclear cfDNA assays further enhances diagnostic performance, thereby underscoring their complementarity. Similar aberrant cf-mtDNA fragmentation patterns have been observed across multiple solid tumors, including liver, breast, and lung cancers¹⁶. Large-scale multi-cancer analyses have indicated that cf-mtDNA fragmentomic signatures not only can discriminate states of cancer vs. health but also can inform tissue-of-origin prediction. In hepatocellular carcinoma, whose early diagnosis remains challenging, cf-mtDNA fragment-level alterations appear early in tumorigenesis and show promise for early detection and disease monitoring¹⁸.

Despite its promising performance in early neoplasia detection, cf-mtDNA fragmentomics faces challenges in specificity. Circulating mtDNA can increase in non-malignant conditions such as tissue injury and sterile inflammation, in which it serves as a damage-associated molecular pattern and correlates with inflammatory mediators. Future studies should benchmark cf-mtDNA signatures against common inflammatory and autoimmune conditions, and prioritize large-scale prospective cohorts to assess false-positive risk and refine decision thresholds.

Collectively, these findings position cf-mtDNA fragmentomics as a robust and scalable strategy for pan-cancer early screening. Capture-based sequencing approaches targeting mtDNA offer high coverage at relatively low cost, and support simultaneous assessment of copy number, mutations, and fragmentation patterns. Although mechanistic understanding and large prospective validation remain ongoing challenges, cf-mtDNA fragmentomics is a compelling frontier poised to reshape noninvasive cancer detection paradigms.

From theory and mechanism to translation

Liquid biopsy research exemplifies the iterative progression from conceptual discovery to mechanistic understanding and ultimately clinical translation. The initial detection of cfDNA in maternal plasma established the foundational concept that pathological tissues can release nucleic acids into circulation⁴. Subsequent mechanistic studies elucidated active and passive DNA release pathways, including exosomal secretion, apoptosis, and necrosis, thus revealing that fragmentation patterns and molecular endpoints contain biologically meaningful information¹⁹. Investigations on virus-associated malignancies, such as Epstein-Barr virus–driven nasopharyngeal carcinoma and hepatitis B virus–associated hepatocellular carcinoma, have further clarified how infectious agents and host immunity shape circulating DNA profiles. More recently, immunogenomic analyses of circulating T- and B-cell receptor repertoires have expanded this framework, by capturing adaptive immune dynamics in response to disease and therapy²⁰. Parallel advances in multi-cancer early detection assays based on ctDNA methylation underscore the feasibility of translating mechanistic insights into scalable clinical applications⁸.

The current central challenge is integrating these diverse insights, spanning ctDNA, global cfDNA features, microbial cfDNA, and cf-mtDNA, into a unified, patient-centric diagnostic paradigm (Figure 2). Achieving this goal will require standardized pre-analytical workflows, analytically validated assays, comprehensive reference datasets, and computational frameworks capable of synthesizing multi-omic signals into clinically actionable outputs. As illustrated by recent advances in cf-mtDNA fragmentomics, bridging molecular innovation with clinical utility is particularly critical for early cancer detection. Successful translation will depend on rigorous prospective validation, regulatory alignment, and cost-effective implementation, with implications extending beyond oncology to immunological, autoimmune, and transplant-related diseases.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

A comprehensive multi-omics liquid biopsy pipeline integrating nuclear, mitochondrial, and microbial sources. This schematic presents a cyclical, four-stage workflow (middle circle), including feature recognition, biomarker discovery, model construction, and clinical evaluation, on the basis of liquid biopsy specimens derived from nuclear, mitochondrial, and microbial origins (inner circle). The outer circle highlights 5 major omics layers (genomics, transcriptomics, proteomics, fragmentomics, and microbiomics) that provide a systems-level understanding of disease processes. Each layer highlights distinct molecular features, such as SNVs, RNA abundance, protein quantification, fragment size, and microbial composition. These integrated molecular profiles collectively support the identification of biomarkers and the development of predictive models for subsequent clinical application. SNVs, single nucleotide variants; CNV, copy number variation; evRNA, extracellular vesicle RNA; PTMs, post-translational modifications.

Future perspectives

As liquid biopsy advances toward broader clinical implementation, the extreme scarcity of tumor-derived DNA within the total cfDNA pool, particularly in early-stage disease, remains a central challenge. This “needle-in-a-haystack” problem is compounded by background cfDNA originating from normal tissue turnover and leukocyte lysis, which can obscure low-frequency tumor signals. Continued progress will therefore require improvements in analytical sensitivity and signal-to-noise discrimination. Technologies such as droplet digital PCR and single-molecule sequencing have substantially advanced rare-variant detection by minimizing amplification bias and background interference. Nevertheless, technical gains alone are insufficient. Pre-analytical variability, including blood collection, processing time, storage conditions, and centrifugation protocols, remains a major source of inconsistency. Therefore, standardized workflows and rigorously enforced quality controls are needed. For example, the choice of blood collection tube (e.g., standard EDTA vs. cfDNA-stabilizing tubes) and delays in plasma processing can increase leukocyte lysis-derived background DNA, thereby altering fragment-size distributions and hindering tumor signal detection²¹. Similarly, differences in centrifugation protocols (e.g., single-spin vs. double-spin) affect residual debris and genomic DNA contamination, thus highlighting the need for standardized workflows. Community efforts are increasingly promoting harmonized practices and quality assurance guidelines for ctDNA testing²².

Beyond technical considerations, economic feasibility is a critical determinant of large-scale use, particularly for population-level screening in asymptomatic individuals. Although sequencing costs continue to decline, comprehensive multi-analyte assays remain resource-intensive. Future solutions will require a balance between analytical depth and cost efficiency, supported by automation, optimized sequencing strategies with reduced read depth requirements, and artificial intelligence–driven bioinformatics pipelines for rapid interpretation and data integration. For multi-cancer early detection tests, population implementation introduces additional considerations beyond analytical performance. Key issues include the effects of false positives, diagnostic burden, cost-effectiveness, resource use, and integration into clinical pathways. Evidence frameworks emphasize the need for well-designed prospective studies that capture detection metrics, clinical harms, diagnostic trajectories, and health outcomes, to enable informed integration into screening programs.

In the future, the field is expected to combine multiple liquid biopsy dimensions, including ctDNA mutations, global cfDNA fragmentomics, immunogenomic signals, microbial cfDNA, and cf-mtDNA, into unified, clinically actionable frameworks. Such integration promises a more holistic and dynamic view of disease biology from a single blood draw. As validation in large prospective cohorts progresses, and regulatory and economic barriers are addressed, liquid biopsy is likely to become a cornerstone of precision medicine, by enabling earlier detection, real-time monitoring, and increasingly proactive disease management in oncology and beyond.

Author contributions

Conceived and designed the paper: Jinliang Xing and Kaixiang Zhou.

Collected and synthesized the literature: Wenjie Guo and Yang Liu.

Contributed data or analysis tools: Haiying Dong and Xiang Li.

Wrote the paper: Wenjie Guo and Yang Liu.

Performed the analysis: Xu Guo and Haitao Guo.

Reviewed and revised the paper: Jinliang Xing and Kaixiang Zhou.