Closing in on cfDNA-based detection and diagnosis

It has been appreciated for some time that analysis of cell-free DNA (cfDNA) in bodily fluids such as plasma (so-called liquid biopsies) might provide a non-invasive way to detect and diagnose cancers earlier and more accurately without the need for invasive and/or costly procedures, such as biopsies and imaging. However, moving this possibility from theory to reality has been hampered by several factors, such as the small amount of cfDNA released by some tumours (especially early-stage tumours) and challenges in distinguishing tumour-specific cfDNA from cfDNA that arises from normal cells.

In a previous paper, Shen et al. reported a methodology termed cell-free methylated DNA immunoprecipitation and high-throughput sequencing (cfMeDIP-seq), which enriches for methylated cfDNA fragments and allows comprehensive profiling of methylated cfDNA to detect and classify a range of tumours. Following on from this, groups led by Daniel De Carvalho, Toni Choueiri, Matthew Freedman and Gelareh Zadeh now report two studies that validate the utility of this technology for early detection of renal cell carcinomas (RCCs) and diagnosis of central nervous system (CNS) tumours.

Credit: Lara Crow/Springer Nature Limited

Nuzzo, Berchuck, Korthauer, Spisak et al. examined RCC, as ~35% of cases are diagnosed at a late stage, but there is no approved test to screen for this disease in the general population. Furthermore, screening of patients with heritable syndromes that predispose to RCC development relies on imaging, which is expensive and exposes patients to radiation.

The authors conducted cfMeDIP-seq on plasma samples from 69 stage I–IV RCC cases (23 were stage I or II) and 13 normal controls. Eighty percent of cases and controls were randomly selected as a training set; within these samples the top 300 differentially methylated regions (DMRs) between the two groups were identified. These were used to build a classifier that then assigned a methylation score to the 20% of samples that were not used to build the classifier, and the entire process was repeated 100 times. This process clearly separated cases from controls, with 67 of 69 cases assigned a higher median methylation score than all controls, giving a mean area under the receiver operating characteristic (AUROC) curve of 0.990 (a value of 1 would indicate perfect classification).

The same process to build a classifier was repeated with the RCC cases and 21 urothelial bladder cancer (UBC) cases and was able to clearly distinguish between these two genitourinary tumour types (mean AUROC curve of 0.979). The authors also conducted the same analysis on cfDNA from urine (30 RCC and 15 control samples). Despite the protocol not being optimized for use on urine samples, it still achieved a mean AUROC curve of 0.858. Urinary detection of genitourinary tumours is perhaps more attractive than plasma-based detection, and the authors state that further refinement of this method is ongoing.

Nassiri, Chakravarthy, Feng et al. analysed the utility of plasma cfMeDIP-seq for discriminating among different intracranial CNS tumours. Accurate diagnosis of these tumours is a clinical challenge; only some require surgical interventions, so accurate classification without the need for invasive biopsies is highly desirable.

The authors conducted cfMeDIP-seq on 59 plasma samples of patients with diffuse gliomas and compared these with previously generated data on 389 patients with extracranial tumours and healthy controls. Using the same procedure as described above, they built a classifier that accurately distinguished gliomas from other cancer types and healthy controls (mean AUROC curve ~0.99). As differences in CNS tumour methylation profiles have been successfully used to discriminate among different CNS tumours, the authors generated cfMeDIP-seq data from a further 161 samples of several different intracranial tumours (many of which have similar cells of origin) including gliomas, meningiomas, hemangiopericytomas and brain metastases. Classifying using the same methodology reliably distinguished most of the tumours from each other, indicating the potential of methylated plasma cfDNA to improve diagnosis of these tumours.

“two studies that validate the utility of this technology for early detection … and diagnosis”

Together, these studies move us a step closer to using methylated cfDNA profiling to accurately classify and detect tumours.


Original articles

  1. Nuzzo, P. V. et al. Detection of renal cell carcinoma using plasma and urine cell-free DNA methylomes. Nat. Med. 26, 1041–1043 (2020)

    CAS  Article  Google Scholar 

  2. Nassiri, F. et al. Detection and discrimination of intracranial tumors using plasma cell-free DNA methylomes. Nat. Med. 26, 1044–1047 (2020)

    CAS  Article  Google Scholar 

Related article

  1. Shen, S. Y. et al. Sensitive tumour detection and classification using plasma cell-free DNA methylomes. Nature 563, 579–583 (2018)

    CAS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Sarah Seton-Rogers.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Seton-Rogers, S. Closing in on cfDNA-based detection and diagnosis. Nat Rev Cancer 20, 481 (2020).

Download citation

Further reading


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing