Liquid Biopsy: How a Blood Test Is Replacing Surgical Cancer Biopsies

What Is Liquid Biopsy and Why Does It Matter?

Cancer diagnosis has long depended on surgical tissue biopsies: a surgeon removes a sample from a suspected tumour, a pathologist examines it under a microscope, and only then can clinicians confirm a malignancy and characterise its molecular profile. The procedure is invasive, painful, and often impractical when tumours are located in the lung, liver, or brain. For patients with multiple metastatic sites, a single biopsy captures only a fraction of the tumour's genetic diversity, leaving clinicians with an incomplete picture of what is driving disease progression.

Liquid biopsy changes this paradigm fundamentally. When cancer cells die or shed material into the bloodstream, they release fragments of their DNA, intact circulating tumour cells, and small RNA molecules called microRNAs. A blood draw of as little as 10 millilitres can capture these biomarkers. Sequencing that material with next-generation sequencing platforms allows oncologists to identify driver mutations, copy number alterations, gene fusions, and methylation patterns associated with specific cancer types, all without touching the tumour directly.

The clinical implications are substantial. A patient undergoing chemotherapy can have their blood tested every few weeks to see whether resistance mutations are emerging, enabling oncologists to switch therapies before visible disease progression occurs on imaging. Patients in remission can be monitored for molecular residual disease, the microscopic cancer signals that precede relapse by months. And in population screening contexts, liquid biopsy offers the tantalising prospect of detecting cancer before symptoms appear, at a stage when treatment outcomes are dramatically better. As the field matures, liquid biopsy is evolving from a research tool into a core component of precision oncology workflows worldwide.

The Biology Behind Circulating Tumour DNA

Circulating tumour DNA, or ctDNA, is the primary analyte in most liquid biopsy assays. Tumours shed DNA into the plasma through apoptosis (programmed cell death), necrosis, and active secretion. These fragments, typically 150 to 200 base pairs in length, circulate freely in plasma and have a half-life of approximately one to two hours, meaning that a liquid biopsy reading reflects very recent tumour biology rather than a historical snapshot. This rapid turnover is actually a clinical advantage: it allows treatment effects to be measured quickly and resistance to be caught early.

The central technical challenge is that ctDNA is a minority species in a sea of normal cell-free DNA released by healthy tissues. In early-stage cancers, ctDNA may comprise less than 0.1 percent of all circulating DNA, sometimes as low as 0.01 percent. Detecting mutations at this frequency demands sequencing technologies with extremely low error rates. Two approaches dominate the field: targeted deep sequencing, which sequences a predefined panel of cancer-relevant genes to enormous depth (often 10,000 to 100,000 times coverage), and whole-genome sequencing at lower depth combined with machine learning analysis of copy number and methylation patterns.

Beyond ctDNA, liquid biopsy platforms also analyse circulating tumour cells, which are intact cancer cells that have detached from the primary tumour and entered circulation. Though rarer than ctDNA fragments, circulating tumour cells can be cultured and subjected to drug sensitivity testing, offering a direct functional readout of tumour behaviour. Exosomal RNA, small vesicles secreted by cancer cells carrying RNA cargo, represents another emerging liquid biopsy analyte that may improve sensitivity for certain tumour types, particularly pancreatic cancer, which is notoriously difficult to detect early.

DNA methylation patterns are emerging as perhaps the most powerful signal for multi-cancer early detection. Different cell types in the body have distinct methylation signatures, and cancer cells frequently display aberrant hypermethylation or hypomethylation at specific genomic loci. By reading these epigenetic patterns from plasma DNA, algorithms can not only flag the presence of cancer but predict the tissue of origin with meaningful accuracy, a critical capability for a test that must tell clinicians where to look when it returns a positive signal.

FDA-Cleared Tests and the Clinical Landscape

The liquid biopsy market has moved from academic research into regulated clinical practice at pace. The US Food and Drug Administration has cleared or approved several liquid biopsy assays for specific clinical indications, establishing a regulatory framework that other countries are following.

Guardant Health's Guardant360 CDx, approved in 2020, was the first FDA-approved companion diagnostic using liquid biopsy, cleared to identify patients with non-small-cell lung cancer who may benefit from osimertinib (Tagrisso). The test panels 74 genes and uses digital sequencing to achieve the sensitivity required for clinical decision-making. Foundation Medicine's FoundationOne Liquid CDx covers 324 genes and is approved as a companion diagnostic for multiple targeted therapies across lung, breast, colorectal, and ovarian cancers. Roche's cobas EGFR Mutation Test v2 remains a standard reference for detecting EGFR mutations in lung cancer patients, with clinical validation from the FASTACT-2 and ENSURE trials.

For early detection, GRAIL's Galleri test has generated the most clinical attention. The test analyses methylation patterns across more than 100,000 genomic regions and claims to detect signals for over 50 cancer types from a single blood draw. The NHS-GRAIL trial in the United Kingdom enrolled 140,000 participants and published interim results in 2023 showing that the test detected cancer at a positive predictive value of 44 percent, meaning roughly 4 in 10 positive tests corresponded to a true cancer, with a cancer signal origin identified correctly in 88 percent of detected cases. While sensitivity for stage I cancers remains below 20 percent in most tumour types, the test shows much higher sensitivity for cancers such as ovarian, pancreatic, and head and neck malignancies that are currently diagnosed late and lack effective screening tools.

This landscape connects directly to the broader work happening in precision oncology and tumour profiling, where comprehensive molecular characterisation of an individual patient's cancer is guiding treatment selection beyond simple histological classification. Liquid biopsy provides a dynamic, repeatable layer of molecular data that complements tissue-based genomic profiling.

Monitoring Treatment Response and Detecting Resistance

One of the most immediately impactful clinical applications of liquid biopsy is serial monitoring of patients on targeted therapy. When a patient with EGFR-mutant lung cancer responds to osimertinib, ctDNA levels fall sharply within weeks of treatment initiation. Conversely, when resistance develops, often through secondary mutations such as EGFR C797S or MET amplification, ctDNA levels rise and the resistance mutation becomes detectable in plasma months before progressive disease appears on CT imaging.

A pivotal demonstration of this principle came from the BFAST (Blood-First Assay Screening Trial), a Phase II/III study in treatment-naive non-small-cell lung cancer patients. The trial used the Foundation Medicine liquid biopsy assay to identify patients with ALK rearrangements from blood alone, comparing outcomes against tissue-based testing. The liquid biopsy-identified ALK-positive cohort showed response rates to alectinib of 87.4 percent, comparable to tissue-identified patients, validating blood-first molecular testing as an equivalent entry point for treatment selection.

In colorectal cancer, the RAS mutation landscape is a critical determinant of whether anti-EGFR antibodies like cetuximab will work. Liquid biopsy can identify RAS mutations emerging under anti-EGFR pressure and can detect the re-emergence of RAS wild-type clones after treatment cessation, informing decisions about rechallenge therapy. The CHRONOS trial, published in Nature Medicine in 2022, demonstrated that ctDNA-guided rechallenge with cetuximab in RAS wild-type colorectal cancer patients yielded meaningful clinical benefit, with liquid biopsy serving as the biomarker gating patient selection.

Molecular residual disease detection after surgery represents another high-value application. In stage II colorectal cancer, landmark work from the DYNAMIC trial showed that ctDNA-guided decisions about adjuvant chemotherapy reduced the number of patients receiving chemotherapy by 15 percent without compromising three-year recurrence-free survival, sparing patients from unnecessary treatment toxicity. This kind of evidence is shifting liquid biopsy from exploratory testing to standard-of-care discussions in major oncology guidelines.

How Artificial Intelligence Unlocks Liquid Biopsy's Full Potential

The data volumes generated by next-generation sequencing of plasma DNA are immense. A single whole-genome sequencing run on a plasma sample generates billions of base reads, each carrying information about mutation status, fragment length, methylation state, and nucleosome positioning. Extracting clinically actionable signals from this data, while suppressing artefacts introduced by sequencing chemistry, PCR amplification, and germline variation, is fundamentally a machine learning problem.

AI approaches applied to liquid biopsy span several technical domains. Error suppression algorithms use neural networks trained on matched tumour-normal pairs to distinguish true somatic mutations from sequencing errors that occur at similar frequencies. Fragment length analysis uses the observation that tumour-derived DNA fragments tend to be shorter than those released by healthy cells; machine learning classifiers trained on these length distributions improve ctDNA detection sensitivity without requiring knowledge of specific mutations. Tissue-of-origin classifiers use methylation patterns across thousands of CpG sites as features, training deep learning models to predict which organ is the source of cancer signals in multi-cancer early detection contexts.

GRAIL's Galleri test uses a machine learning architecture trained on data from over 6,600 cancer cases and 3,000 cancer-free individuals. The model integrates methylation signals across 104,000 targeted regions to produce both a cancer signal detected/not detected output and a cancer signal origin prediction. The tissue-of-origin accuracy of 88 percent in confirmed cancer cases is substantially higher than what rule-based analysis of methylation patterns can achieve, demonstrating the direct value of machine learning in this clinical domain.

Multimodal AI integration represents the next frontier. Researchers at the University of Cambridge and elsewhere are building models that combine ctDNA data with protein biomarkers, imaging features, and electronic health record data to produce integrated cancer risk scores. This mirrors the broader trend in AI transforming medical diagnosis, where the greatest gains come not from any single data modality but from intelligent fusion of complementary information streams. For liquid biopsy specifically, combining ctDNA with CA-125 (ovarian cancer), AFP (liver cancer), or PSA (prostate cancer) protein markers has been shown to increase sensitivity for early-stage disease by 10 to 20 percentage points compared to either biomarker category alone.

The connection to genomic medicine runs deep. Understanding which mutations liquid biopsy detects requires fluency in the genomic landscape of cancer, a domain explored in depth in AI and genomics: machine learning applied to DNA. Mutations detected in ctDNA can also inform pharmacogenomic decisions about drug dosing and toxicity risk, a topic covered comprehensively in pharmacogenomics explained.

Limitations, Challenges, and the Road Ahead

Liquid biopsy is a powerful technology with real and acknowledged limitations that the field continues to work through. Sensitivity for early-stage cancers remains the most significant clinical gap. Stage I cancers shed very low quantities of ctDNA, and in some tumour types such as glioblastoma, the blood-brain barrier substantially restricts the passage of tumour-derived DNA into systemic circulation, reducing test sensitivity precisely where early detection would be most valuable.

Clonal haematopoiesis of indeterminate potential, or CHIP, represents a confounding biological phenomenon. As people age, haematopoietic stem cells accumulate somatic mutations that expand clonally within the blood. These mutations, in genes such as DNMT3A, TET2, and ASXL1, are detectable in plasma DNA and can be misattributed as tumour-derived signals. Careful filtering against white blood cell DNA is required to exclude CHIP variants from liquid biopsy cancer calls, adding a layer of analytical complexity and cost.

Cost and reimbursement remain barriers to widespread adoption. Comprehensive liquid biopsy panels currently cost between $1,500 and $5,000 per test in the United States. Insurance coverage is improving but remains inconsistent, particularly for monitoring indications beyond initial diagnosis. For population-level early detection screening to be viable, per-test costs will need to fall substantially, a trajectory that improving sequencing economics and competitive market dynamics are likely to deliver over the next decade.

Standardisation of pre-analytical variables is another active area of work. The way blood samples are collected, how quickly plasma is separated, and how samples are stored before analysis can all affect ctDNA recovery and introduce variability between studies and clinical sites. Bodies including the International Liquid Biopsy Standardization Alliance are working to establish consensus protocols that will make results comparable across laboratories and healthcare systems.

Despite these challenges, the trajectory of liquid biopsy development is unmistakably positive. The PATHFINDER study, the SUMMIT trial, the NHS-GRAIL trial, and dozens of ongoing prospective studies are accumulating the clinical evidence base that regulators and payers require to expand reimbursement and embed liquid biopsy into standard screening guidelines. Within the broader framework of precision oncology, liquid biopsy is becoming the dynamic molecular layer that connects initial tumour characterisation to real-time treatment monitoring, transforming cancer management from episodic intervention to continuous molecular surveillance.

Summary: A New Standard for Cancer Detection and Monitoring

Liquid biopsy has transitioned from a scientific concept to a clinically validated tool in less than a decade. FDA-cleared assays now guide treatment selection across multiple tumour types. Serial monitoring with ctDNA is changing how oncologists manage resistance to targeted therapies. Multi-cancer early detection tests are enrolling hundreds of thousands of participants in prospective trials that will define how population screening operates in the next decade.

The technology's power scales with the sophistication of the data analysis applied to it. Machine learning has been the critical enabler, transforming raw sequencing reads into clinically interpretable signals at the detection limits that early-stage cancer demands. As AI capabilities advance and sequencing costs fall, liquid biopsy will expand its role from a specialised oncology tool to a routine component of preventive healthcare, capable of detecting cancer at its most treatable stages in patients who feel entirely well.

For clinicians, patients, and health systems, liquid biopsy represents one of the clearest demonstrations that precision medicine is not a theoretical aspiration but an operational reality reshaping cancer outcomes today. The central promise of the field, catching cancer earlier, monitoring it more dynamically, and personalising treatment with molecular precision, is being fulfilled in clinical practice right now, with more capability arriving each year.