MCED: One for All, All for One

Aptly named, multi-cancer early detection (MCED) aims to identify multiple cancer types as early as stage I with a singular test. MCED has the potential to revolutionize cancer diagnostics – but there is plenty of work to do. Here, we present a collection of research stories that showcase the fast-moving field of MCED. 

Urine luck
 

Most MCED tests have focused on circulating cell-free DNA (cfDNA) to identify a tumor’s site of origin; however, DNA-based methods are not without limitations, including their limited sensitivity to stage I cancers. Furthermore, MCED platforms that use cfDNA struggle to detect certain cancer types – particularly, genitourinary and brain. Considering these challenges, a group of international researchers turned to cancer metabolism (1), exploring whether free glycosaminoglycans (GAGomes) from the urine and blood plasma could act as metabolic cancer biomarkers for MCED.

The study took 2,064 samples from 1,260 participants and tested for 14 different cancer types. Three machine learning models were developed based on urine and plasma GAGomes – and a clear biological association between free GAGomes and the occurrence of any type of cancer was observed. All 14 cancers were detected, and the study further revealed that GAGomes have the potential to detect any cancer with up to 62 percent sensitivity to stage I disease at 95 percent specificity.

Hot new platform 
 

Though methylation profiling in cell-free DNA for MCED is popular for its ability to enable broad characterization of cancer-associated patterns, these assays are unfeasible because of the high cost and labor-intensive nature of whole-genome sequencing. Researchers in Singapore leveraged the fact that cancer-specific signatures are often dense in the CpG sites of the genome sequence, and developed the Heatrich-BS assay – a technique that uses thermal denaturation to isolate CpG-rich regions for undivided, methylation profiling (2).

The assay was paired with a bioinformatics algorithm that estimated cfDNA global tumor fraction from low-depth sequencing. To validate the performance of Heatrich-BS, samples of cfDNA were obtained from five healthy volunteers and 15 patients with colorectal cancer (CRC). The assay successfully measured tumor fractions with minimal sequencing efforts. 

Further, Heatrich-BS was assessed on its ability to monitor cancer progression. From an alternative cohort of CRC patients, longitudinal CEA measurements and CT scans were obtained for benchmarking tumor fraction predictions by Heatrich-BS. The team discovered that the assay was more sensitive in the longitudinal monitoring of patients with cancer than existing CEA protein biomarker assays in the detection of CRC at low tumor fractions.

Overall, the study reports a universally affordable (US$30), non invasive, and highly sensitive sequencing assay with a quick turnover time. In the future – researchers hope it can be fully integrated into point-of-care settings to enable scalable implementation of cfDNA methylation profiling in liquid biopsies.

Top of the classifier 
 

The fresh possibilities that MCED tests can offer for cancer patients is exciting; however, until now, there have been no rigorous and systematic comparisons of genomic features from cfDNA for MCED testing. In a recent study, researchers used the clinical limit of detection (LOD) – based on circulating tumor allele fraction (cTAF) – as a benchmark to compare sensitivities of different approaches (3).

A group of 2,800 subjects – 1,628 with cancer and 1,172 without  – took part in the first circulating cell-free genome atlas (CCGA) substudy and were randomly assigned to validation sets. After taking blood samples, cfDNA was extracted. Ten different, machine-learning classifiers were automated to detect a cancer signal and independently validated. The team revealed that the best-performing classifiers – when evaluated at 98 percent specificity – were whole-genome methylation, single nucleotide variant with paired white blood cell background removal, and the pan-feature classifier. 

The researchers concluded that methylation was optimal for cfDNA MCED tests – and the most promising genomic feature for cancer signal detection. Further, it was revealed that cTAF was a fundamental predictor of classifier performance.