Investors keep the faith in cancer liquid biopsies

Despite slow uptake in the clinic, liquid biopsy tests incorporating methylation and other signatures are showing increasingly compelling data for non-invasive detection of cancer.
Dublin, Ireland

Search for this author in:

Liquid-biopsy platforms, which have long faced considerable scepticism as to their utility in clinical practice, are attracting a new bout of intense investor interest.

A small blood sample can screen for cancer by detecting DNA shed by tumors into circulation.Credit: Kong Hon Loh / Alamy Stock Photo

Adaptive Biotechnologies completed a hefty $345 million initial public offering in July; Freenome raised $160 million series B funding in the same month; and Thrive Earlier Detection launched with a $110 million series A round in May. Gleaning meaningful insights from tiny amounts of circulating tumor DNA (ctDNA), the premise underpinning liquid biopsy tests, has found some use in complementing confirmed cancer diagnoses. Companies with commercial offerings, like Adaptive and Guardant Health, as well as numerous academic investigators, have long touted the potential of tests run on a simple blood draw to aid prognosis, estimate tumor size, guide treatment selection, monitor resistance to therapy, and detect disease recurrence; however, adoption in clinical practice remains low. And use in population screening is an altogether more complex problem, given the very high signal-to-noise ratio in that setting.

Tumor-derived DNA is just a small percentage of the total cell-free DNA fraction in circulation. Yet the promise of early detection remains alluring, given the persistently wide gap in outcomes between patients who receive a cancer diagnosis sooner and those who receive one later. “Most cancers—not all—will be cured by earlier detection,” says Justin Stebbing, professor of cancer medicine and oncology at Imperial College London.

The fact that ctDNA is present in patients’ blood has been recognized for decades. But it is only with next-generation sequencing (NGS) technologies, which can distinguish low levels of ctDNA in the total cell-free DNA shed by all cells—whether cancerous or not—that it has become possible to consider it a potentially useful analyte for diagnosis. Whether NGS-based assays can do so with the sensitivity (correct identification of those with cancer) and specificity (correct identification of those who are cancer free) required for a clinically useful screening test remains an open question. Liquid biopsy testing, even in late-stage patients, remains a minority pursuit. “It’s still a massive work in progress,” says Stebbing. “In the vast majority of patients, it’s not been incorporated into clinical practice.”

The first prospective studies for early cancer detection with liquid biopsies are underway in average-risk and high-risk individuals—as opposed to individuals who already have a confirmed cancer diagnosis. It will still, however, take several more years for the data to mature sufficiently to the point where blood-based tests can be properly assessed for their clinical utility. In the meantime, leading companies in this space continue to pursue key questions as they hone their products: which biological signals to look for, how to interpret the results, and which indications to focus on. As competition in the liquid biopsy space heats up, the various contenders seek to differentiate themselves across all of these parameters (Table 1).

Grail, which was spun out from NGS technology developer Illumina in 2016 and galvanized the entire field with a dramatic $900 million in equity raised in 2017—it has raised $1.5 billion in total—has largely abandoned searching for cancer-associated mutations in ctDNA in favor of studying its methylation signatures instead. A recent finding that even non-malignant hematopoietic cells can accumulate clonal mutations complicates the interpretation of genotyping cell-free DNA, says chief scientific officer and cofounder Alex Aravanis. In any case, he maintains that the cell-free DNA methylome provides a richer signal than cancer-associated mutations do. For a particular region of the genome being investigated, the DNA would typically harbor just two to three mutations of interest, whereas methylation tags are abundant. “You might have hundreds or thousands of methylation changes in the same segment of the genome,” he says. Grail’s sequencing database of cancer and non-cancer methylation signatures covers about 30 million methylation sites across the genome. As well as indicating the presence—or absence—of cancer, methylation signatures also provide insights into the tissue of origin of any cancer that is detected, which is important for subsequent clinical follow-up.

At the American Society of Clinical Oncology meeting earlier in the summer, Grail and its clinical collaborators reported detection rates for its test that ranged from 59% to 86% (at 99% specificity) across twelve prespecified cancer types. Combining all cancer types but slicing the results by stage yielded 34%, 77%, 84% and 92% for stage I to IV cancers, respectively. “We think it more than clears the bar for a clinically meaningful test,” Aravanis says. “We’re now ready to bring this to the clinic.” The company, which has from the outset focused on detecting a wide range of cancers, will start a prospective study next year to test its multicancer liquid biopsy diagnostic. For the first time, participants will receive their test results and will, if necessary, be in a position to act on them.

Thrive Earlier Detection is also tackling multiple indications from the start. “We are aiming ultimately to build earlier detection into routine care for what we call average-risk individuals,” says Thrive CEO Steven Kafka. “This needs to be and can be, given the technology, a multicancer approach.” The recently launched company is developing a liquid biopsy platform based on CancerSEEK, a prototype test developed by Nickolas Papadopoulos, Kenneth Kinzler and Bert Vogelstein of the Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins University. CancerSEEK is a multi-analyte test that recognizes mutations at 1,933 distinct genetic loci, as well as a panel of validated protein biomarkers of cancer.

Early results with CancerSEEK may explain investors’ enthusiasm. In a retrospective study in 1,005 patients with confirmed, non-metastatic cancer (stages I–III), it achieved detection rates from 69–98% (at >99% specificity) across five cancers (ovary, liver, stomach, pancreas and esophagus) for which there are no screening tests available. A machine-learning-based algorithm localized the tumors to two anatomic sites in a median 83% of patients and to a single organ in a median 63% of patients. CancerSEEK’s median sensitivity was 43% for stage I cancers, 73% for stage II cancers and 78% for stage III cancers. In the absence of any screening tests for most types of cancer, that level of sensitivity may be sufficiently useful for detecting cancer earlier than is now the case. “It’s important we say ‘earlier’ as opposed to ‘early’,” Kafka says. “It isn't just a function of finding cancer at stage I, although it would be great if we could do that.”

Thrive is now collaborating with Johns Hopkins and the healthcare provider Geisinger on a five-year prospective screening study, DETECT, which has enrolled 10,000 healthy female participants ages 65–75. An interim data read-out, based on one year of follow-up, is expected in the middle of next year.

Freenome is casting the net more widely than its rivals in the hunt for early signs of cancer. Its ‘multi-omic’ approach looks at cell-free DNA, methylation signatures and proteins, both from nascent tumors and from the immune system. It then integrates these signals using artificial-intelligence-driven analytics. The foundation for Freenome’s approach is that circulating cells of the immune system are detecting and responding to cancer in its earliest stages; these are ignored by most other companies, which focus their liquid biopsies on the ctDNA fraction shed exclusively by tumors. “They ignore the rest of the 99-plus percent of the cell-free DNA,” says chief scientific officer Jimmy Lin. In one publicly disclosed retrospective study, Freenome obtained 85% mean sensitivity (at 85% specificity) for stage I and stage II colorectal cancers testing cell-free DNA from tumor and immune cells, without including any other analytes. “We believe we can well exceed that,” says Freenome cofounder and chief partnerships officer Charles Roberts.

The multimodal approach is “valid,” says Eleftherios Diamandis, head of clinical biochemistry at Mount Sinai Hospital and University Health Network in Toronto and at the University of Toronto, but he adds a caveat: “As you increase the number of analytes you measure, you increase the sensitivity—you detect more cancers—but you may decrease the specificity at the same time,” he says. Freenome’s Lin counters that its proprietary machine-learning approach to interpreting the data its test generate can avoid this problem.

Freenome is taking a stepwise approach to develop its platform, with an initial focus on colorectal cancer. The choice of indication reflects the fact that colorectal cancer already has a well-defined screening test, Exact Sciences’ Cologuard stool DNA test, against which Freenome can pit its own test. A validation study called AI-Emerge, which involves 3,400 participants, including healthy volunteers undergoing routine colonoscopy and people with newly diagnosed colorectal cancer, is underway and will read out early next year. Freenome is not alone in the colorectal cancer space. Guardant Health, an early leader in liquid biopsy testing for treatment selection, is also developing an early detection test for colorectal cancer, Lunar-2, which combines sequence and epigenomic analyses of ctDNA, as well as a bioinformatics classification system for eliminating benign clonal mutations arising during hematopoiesis. “Later in the second half of 2019, we’re going to start our pivotal prospective screening study in colorectal cancer,” says AmirAli Talasaz, Guardant’s cofounder, president and chief operating officer. This will recruit 10,000 participants with average risk of developing colorectal cancer.

At present, population-based cancer screening is very much a hit-and-miss affair. Computed tomography, colonoscopy, Pap testing and mammography have demonstrated varying degrees of benefit in reducing deaths from lung cancer, colorectal cancer, cervical cancer and breast cancer, respectively, but the continued mortality from those conditions underlines the deficiencies of these methods. For many other types of cancer, including pancreatic cancer and liver cancer, there are no effective screening tests. Liquid biopsy developers have plenty of scope, therefore, to bring a whole new approach to cancer screening. But developing an understanding of the clinical significance of a test result will be every bit as important as developing an accurate test. “We know already we can detect cancer early,” says Nitzan Rosenfeld, cofounder and chief scientific officer at Cambridge, UK-based Inivata, pointing to the identification of maternal cancers during cell freeDNA-based non-invasive prenatal testing for aneuploidy (abnormal numbers of chromosomes) during pregnancy. “The hard part is not whether it’s possible, but whether it’s clinically useful or not,” he says. Diamandis concurs: “Screening may lead to overdiagnosis, which may lead to overtreatment,” he says. But, Stebbing notes, managing this problem is an intrinsic aspect of cancer care. “We have problems with overdiagnosis already, with mammography.” Underdiagnosis of early-stage cancers is still a much bigger problem.

Table 1 | Selected liquid biopsy tests for cancer





Adaptive Biotechnologies


Test employing multiplex polymerase chain reaction (PCR) and next-generation sequencing (NGS) for detecting minimal residual disease in acute lymphoblastic leukemia or multiple myeloma

FDA approval 28 September 2018

Epic Sciences, Genomic Health

Oncotype DX AR-V7 Nucleus Detect

Test that detects AR-V7 protein in the nucleus of circulating tumor cells to guide treatment selection for patients with metastatic castration-resistant prostate cancer

Clinical Laboratory Improvement Amendments (CLIA)-certified test

Epigenomics (Berlin)

Epi proColon

Real-time PCR test for methylated cytosine residues in the SEPTIN9 gene for detection of colon cancer

FDA approval 13 April 2016

Foundation Medicine

FoundationOne Liquid

Hybrid-capture-based NGS test that detects clinically relevant genomic alterations (substitutions, insertion/deletions, copy number alterations & selected genetic rearrangements) in 70 oncogenes

FDA breakthrough device designation 28 April 2018


Freenome multianalyte test for colon cancer

‘Multi-omic’ test that integrates biological signals from tumors and the immune system for early detection of colorectal cancer

Research stage


Multicancer early detection test

NGS blood test for detecting multiple cancer types by analyzing ctDNA methylation patterns

FDA breakthrough device designation 13 May 2019

Guardant Health

Guardant360; Guardant Omni; Lunar-1; Lunar-2

Commercially available ctDNA test of a 73-gene panel to guide treatment selection in non-small-cell lung carcinoma (NSCLC) (Guardant360); commercially available 500-gene panel for drug development (Guardant Omni); research-use-only test for detecting minimal residual disease and disease recurrence (Lunar-1); development-stage test for early detection of cancer (Lunar-2)

Guardant360, a lab-developed test, received FDA breakthrough device designation 15 February 2018; the other tests are in development

Inivata (Cambridge, UK)


ctDNA blood test employing tagged amplicon sequencing for detecting genomic alterations in a 36-gene panel, including 8 actionable genes relevant to NSCLC

Lab-developed test with Medicare reimbursement

Mirxes (Singapore)


Quantitative PCR (qPCR) assay for a panel of serum microRNA (miRNA) biomarkers of gastric cancer and breast cancer

The Singapore Health Sciences Authority approval 9 May 2019 for gastric cancer; research stage for breast cancer

Oncimmune Holdings (Nottingham, UK)


ELISA that detects seven autoantibodies directed against tumor-associated antigens in plasma

Available worldwide apart from the US; European Union CE Mark 31 May 2018

Resolution Bioscience

Resolution HRD

NGS assay that detects sequence variations in genes associated with homologous recombination deficiency

FDA breakthrough device designation 30 May 2019

Roche Diagnostics

Cobas EGFR Mutation Test v2

Real-time PCR test that detects 42 defined mutations in the epidermal growth factor receptor (EGFR) gene to guide treatment selection in NSCLC

First FDA approval 1 June 2016

Thrive Earlier Detection


Multianalyte test that combines multiplexed PCR detection of mutations in ctDNA at 1,933 loci with measurements of validated protein biomarkers to diagnose eight common cancer types

Received FDA breakthrough designation 8 August 2018 for detection of mutations and proteins associated with pancreatic and ovarian cancers

Nature Biotechnology 37, 972-974 (2019)

doi: 10.1038/d41587-019-00022-7

Updates & Corrections

  • Correction 14 August 2019: An earlier version of this article omitted mention in Table 1 of Guardant360’s FDA breakthrough device designation 15 February 2018

  • Correction 21 August 2019: The Table 1 incorrectly stated that MirXes’ products were at research stage when GastroClear has been available in Singapore since its approval on 9 May 2019 by the Singapore by the Health Sciences Authority.

Nature Briefing

An essential round-up of science news, opinion and analysis, delivered to your inbox every weekday.