A new approach to liquid biopsy correctly detected the presence of cancer in the majority of cases, and identified the tissue of origin approximately 75% of the time. The test relies on the characteristic fragmentation pattern of circulating tumor DNA (ctDNA, a type of cell-free DNA, or cfDNA), rather than seeking out specific mutations or methylation patterns, as other tests have done.
The test, which was described in Nature,1 was developed by Johns Hopkins Kimmel Cancer Center in collaboration with groups in Denmark and the Netherlands. They call the test DELFI, for “DNA evaluation of fragments for early interception.” Using a machine learning method, the researchers taught DELFI to recognize the fragmentation patterns that suggest the presence of cancer.
Applying the test to 236 cancer patients and 245 healthy individuals, they correctly identified 7 different cancers with a sensitivity of 73% at 98% specificity. The majority of the patients (183 of 236) had stage I to stage III resectable disease, and DELFI positivity identified 72% of those early-stage cancers. The patients had breast, colorectal, lung, ovarian, pancreatic, gastric, or bile duct cancer.
“This could be a breakthrough paper,” said Eleftherios Diamandis, MD, PhD, head of clinical biochemistry at Mount Sinai Hospital, Toronto, Canada, who was not involved in the study. “This paper is adding a new dimension to the analysis.”
Unlike tests that look for specific mutations or changes in methylation patterns, DELFI exploits the chaotic nature of cancer cells to detect the disease. “The nuclei of cancer cells are like a disorganized suitcase, with items from across the genome thrown in haphazardly,” said Victor Velculescu, MD, PhD, professor of oncology at Johns Hopkins Kimmel Cancer Center, in a press briefing. Not all cell-free DNA comes from tumors. To develop the test, the researchers trained DELFI to distinguish the orderly genomes of healthy cells from the jumbled DNA of cancer cells. While mutations and methylation can vary from cancer to cancer, the disorganized genome seems to be universal to all cancers.