A liquid biopsy that combines Illumina’s ‘ultradeep’ sequencing with Grail’s machine learning

Back in 2016, the FDA ushered in a new era by approving the first blood-based cancer-detection test, known as a “liquid biopsy,” which was made by Roche to detect EGFR mutations in non-small cell lung cancer (NSCLC). The ability to find EGFR mutations and other gene irregularities is important, because there are several targeted cancer treatments on the market that can extend survival in lung cancer patients. But these diagnostics are far from perfect.

Now a consortium of scientists from three top U.S. cancer research institutions has developed a new assay that they say could detect multiple targetable lung cancer mutations with a high degree of accuracy.

Some liquid biopsy technologies work by detecting DNA fragments that tumors release into the bloodstream—known as plasma cell-free DNA (cfDNA)—to identify oncogenic drivers and help guide treatment selection. Proponents of liquid biopsies say the technique has to potential to replace invasive tissue biopsies. Problem is, tumors shed much smaller amounts of cfDNA as noncancerous cells do, making it hard to detect the genetic signals that oncologists need to select treatments. In fact, a recent joint review by the American Society of Clinical oncology and College of American Pathologists concluded that it’s still early to adopt cfDNA analysis for routine clinical use.

In an effort to improve the technology, researchers from Memorial Sloan Kettering Cancer Center, MD Anderson Cancer Center and Dana-Farber Cancer Institute have joined forces with DNA sequencing giant Illumina and its liquid-biopsy spinoff Grail to develop a new assay. According to results published in the journal Annals of Oncology, they've come up with a method that can detect mutations in NSCLC patients’ blood with high sensitivity. In some cases, the technology was useful when tissue biopsies were inadequate for analysis.

The new tool uses Illumina’s “ultradeep next-generation sequencing,” which involves reading a region of DNA 50,000 times, on average, to detect low-frequency variants. White blood cells were also sequenced to filter out "clonal hematopoiesis," which are noncancerous signals that can come from bone marrow. The sequencing information was then fed to a machine learning algorithm developed by Grail to determine mutation readouts.

RELATED: #ASCO18: Grail’s first data shows 38%-51% detection of early-stage lung cancer

The researchers tested the assay in 127 patients with advanced, metastatic NSCLC. Among 91 patients whose cancer-driving mutations had been identified from tissue biopsies, the liquid biopsy detected 68, “giving a true positive rate of 75%,” said Bob Li, the medical oncologist at MSK who led the research, in a statement. Among the 19 patients without mutations identified via tissue biopsy, the liquid biopsy also did not detect any irregularities, "meaning there were no false positives and giving a true negative rate of 100%.”

They also tested the technique in a third group of 17 patients whose tumor tissues were deemed insufficient for genotyping. The liquid biopsy identified KRAS mutations in four of those patients. In one patient, the result was confirmed with a tissue biopsy. (In two of the remaining 13 patients, mutations were later found using tissue biopsies.)

Scientists have been exploring new ways to enhance the sensitivity of liquid biopsies. For example, in a recent study published in the journal Science Translational Medicine, a University of Cambridge team used differences in fragment lengths between circulating tumor DNA and noncancerous DNA to tell them apart.

Grail is making progress in its quest to develop highly sensitive blood tests for the early detection of many types of cancer. At the American Society of Clinical Oncology (ASCO) conference last June, the company presented data from its Circulating Cell-free Genome Atlas (CCGA) project, which showed detection rates ranging from 59% to 92% in patients with adenocarcinoma, squamous cell and small cell lung cancers. The rate of false positives—a major concern for the oncology community—was under 2%.

MSK's Li suggests that the quick turnaround time of liquid biopsies makes the technology a potentially valuable addition to the oncologist's toolbox. The 75% detection rate shown in this study indicates that a tissue biopsy might still be needed, he said, but a liquid biopsy could be performed first in the interests of developing a treatment plan earlier. Liquid biopsies could also be a valuable substitution when traditional tissue biopsies are not feasible, he added. “Our results suggest that liquid biopsy can play a complementary role to tissue biopsy in the treatment of lung cancers,” Li said.