November 7, 2024 | A next-generation liquid biopsy approach combining cell-free DNA (cfDNA) “fragmentome” and protein biomarker analyses could aid in the early detection of ovarian cancer, suggests a recent study by researchers in the U.S., The Netherlands, and Denmark. The DELFI-Pro classifier was found to have a high positive predictive value, a key accuracy metric. It could be developed into an accessible method for population-based screening, according to Victor E. Velculescu, M.D., Ph.D., professor of oncology and co-director of the cancer genetics and epigenetics program at the Johns Hopkins Kimmel Cancer Center.
The DELFI acronym signifies the key process—DNA evaluation of fragments for early interception—based on the realization that tiny bits of DNA in the blood, though measuring only about 170 letters in length, come in different sizes and amounts that reflect where they were packaged, Velculescu explains. These data, together with concentration levels of two common protein biomarkers, cancer antigen 125 (CA-125) and human epididymis protein 4 (HE4), are then analyzed using machine learning (a form of artificial intelligence).
CA-125 and HE4 are not approved for clinical use, but are well-known biomarkers utilized in research, he adds. On their own, they cannot reliably detect ovarian cancer, although some clinicians may use them to monitor patient response to therapy.
The same genome-wide cfDNA fragmentomics-based approach is being tested across a variety of cancer types, albeit using different algorithms. It underpins the recently clinically validated laboratory-developed test, called FirstLook Lung, for the early detection of lung cancer that is already certified through the Clinical Laboratory Improvement Amendments (CLIA) program. FirstLook Lung was developed and commercialized by Hopkins spinoff DELFI Diagnostics, founded in 2018 by Velculescu, who serves as board director.
FirstLook Lung is currently the only marketed test being performed out of DELFI Diagnostics’ Palo Alto CLIA lab. The company has recently made several announcements, including collaborations with OSF HealthCare, City of Hope, and Allegheny Health Network to utilize the tool to improve abysmal lung cancer screening rates. “Only 5% of the population is currently getting screened for lung cancer,” says Velculescu. “It’s a tragedy.”
A new research assay called DELFI-TF is being used for monitoring patients during therapy, which is work DELFI Diagnostics is doing with other pharma and biotech partners, Velculescu says. This means that the approach could find utility across the “continuum of cancer care and not just for screening.”
DELFI Diagnostics received its first funding in 2019, the same year the first study describing the approach across seven different cancers appeared in Nature (DOI: 10.1038/s41586-019-1272-6). It wasn’t until then that it was considered possible to use tiny bits of DNA that vary throughout the genome to produce a test for detecting specific cancers. Researchers with the company have since been involved with many other rapidly evolving studies describing different aspects and applications of the fragmentation technique for cancer detection.
Among these is a 2021 study in Nature Communications (DOI: 10.1038/s41467-021-24994-w) where cfDNA fragmentation profiles were used as a prototype for the detection of early-stage lung cancer. Last year, in Cancer Discovery (DOI: 10.1158/2159-8290.CD-22-0659), Velculescu and colleagues showed that the method could be used for early detection of liver cancer, where fragmentation changes were shown to reflect genomic and chromatin changes seen with the disease.
This year, in addition to the latest study specific to ovarian cancer (Cancer Discovery, DOI: 10.1158/2159-8290.CD-24-0393), the research team demonstrated how the fragmentome approach could be used to evaluate changes to repeats of DNA sequences in the genome (“dark matter”) and thereby predict disease in patients with early-stage lung or liver cancer (Science Translational Medicine, DOI: 10.1126/scitranslmed.adj9283). The prospective case-control clinical validation study for FirstLook was also published (Cancer Discovery, DOI: 10.1158/2159-8290.CD-24-0519), showing high performance for detection of lung cancer across diverse individuals eligible for screening in the U.S., says Velculescu.
The application of machine learning to human knowledge of the genome and fragmentome features in the blood for early cancer detection is analogous to the flurry of explorations in the Space Age that enabled a staggering array of technological improvements, he says. The DELFI approach has the potential to lead to a multitude of single cancer tests as well as multi-cancer tests, that could have a major impact on cancer morbidity and mortality worldwide.
Scientists on the research team include veterans of the cancer genomics field responsible for many of the early discoveries about how cancers differ from one another and how they might be better targeted. What no one had, until recently, was any liquid biopsy tool that could be used for broadly accessible screening purposes, says Velculescu, either because the available options were too expensive or insufficiently sensitive to be scaled at the population level.
They ultimately realized they could exploit emerging knowledge that “different parts of the genome are fragmented in different ways” and, depending on the tissue of origin, also have different biological properties, he continues. In cancer cells, DNA becomes disorganized and so, when those cells die and break apart, the DNA fragments left behind are irregular and chaotic.
The bioinformatics experts in the group, including DELFI Diagnostics cofounder and head of data science Rob Scharpf, Ph.D., who is both an associate professor of oncology and biostatistics, and Stephen Cristiano, Ph.D., pointed to machine learning as the obvious way to take advantage of the multitude of data that was going to need interpreting.
For the latest DELFI-Pro liquid biopsy test, “it’s helpful to remember just how terrible ovarian cancer is,” Velculescu says. “The outcome for patients in advanced stages is very poor.” For cancer in general, but especially the deadliest cancers, early detection and intervention is widely recognized as the best hope for reducing mortality rates.
With the latest study, DELFI-Pro has taken “an exciting step in that direction” by outperforming CA-125 in terms of detection—72%, 69%, 87%, and 100% of ovarian cancers for stages I–IV versus 34%, 62%, 63%, and 100% at the same (greater than 99%) specificity, he reports, with similarly good results when used on a separate, smaller set of women. It is critical that people not be incorrectly diagnosed with ovarian cancer because the next step is often exploratory surgery.
The test’s high specificity helps bolster its positive predictive value by reducing the likelihood of a false alarm. Next steps include demonstrating this in larger studies, says Velculescu. “We are working with partners around the world and the U.S. to extend the DELFI analyses for ovarian cancer early detection, looking at existing sample collections of ovarian cancer patients as well as [conducting] new prospective trials to evaluate the [DELFI-Pro] biomarker.”