NEW YORK – A team led by researchers at the University of Southern California has developed a DNA methylation-based assay aiming to classify patients with high-grade serous ovarian carcinoma at an earlier stage.
The team's method, described in a paper published earlier this month in Clinical Cancer Research, utilizes cell-free DNA methylation and machine learning to distinguish between high-grade serous ovarian cancer (HGSOC) and benign masses. The team first used reduced representation bisulfite sequencing to determine the differentially methylated regions between HGSOC and benign ovarian and fallopian tube tissue to guide what features it would use in the test, said Bodour Salhia, an associate professor and interim chair of translational genomics and the head of the lab where the test was developed.
Using both ovarian and fallopian tissue to develop the test, called OvaPrint, was essential as it likely boosted the sensitivity and specificity of the test, she noted. Although ovarian cancers, particularly aggressive cancers, arise from the fallopian tubes, benign masses arise from ovarian tissue, and to distinguish between them, both types of tissue needed to be used. "I think if we were just trying to find differences between fallopian tube and tumor, we may not be as good at picking up the benign mass differences," she said.
According to the CCR paper, the team also used a majority of stage I samples in its discovery process to ensure the test would be effective for early detection.
After conducting a differential methylation analysis, the team identified regions that it wanted to use for the test. They then performed the assay on a largely independent cohort of samples to create a machine-learning classifier. The classifier provides a positive or negative result but can be tiered to indicate whether the risk of having cancer is high, intermediate, or low.
The CCR study included 372 tissue and blood samples, and the test had a positive predictive value of 95 percent and a negative predictive value of 88 percent in differentiating between high-grade ovarian cancer and benign masses.
Salhia noted that the researchers focused on HGSOCs because they're the most frequent and lethal type of epithelial ovarian cancer. They assumed that those cancers would be different than other types of epithelial ovarian cancers and wanted to avoid "bit[ing] everything off in one shot."
When the team's researchers also tested how generalizable the test was for other types of ovarian cancers, they found the performance dropped, she said.
However, the team did find that a high risk score from OvaPrint could indicate nefarious low-grade or borderline tumors that would need more aggressive treatment, which it is looking to explore as the test is developed further. The researchers are also looking to add more features to the test to detect more ovarian cancer subtypes in the future, she added, and the team has received a grant to study other histologic subtypes.
In the CCR paper, the researchers noted that previous screening and detection studies for other tests focused on advanced-stage cancers and grouped all epithelial ovarian cancers together, limiting their scope. The limitations "made it more difficult to discriminate what are, biologically, very different types of tumors," they wrote. OvaPrint bypasses these issues by focusing solely on HGSOC.
The researchers used DNA methylation because it is "a stable and robust covalent mark on DNA," and alterations tend to be more frequent than single point mutations, Salhia said. DNA methylation also provides larger regions of detection and tissue-specific signatures, she noted. And for early detection, a mutation "will not suffice because you will not know the tissue of origin."
The test is intended for patients who are symptomatic and have a mass, whether they are located at a gynecologic clinic, a general practitioner's office, or a cancer center, Salhia noted, and gynecologists will likely be the test's largest end user. Eventually, Salhia said, she'd like to see the test be included in annual routine doctors' visits. The test could also work for cancer screening, since the test is looking for cancer that hasn't been discovered yet but would need to be validated in an asymptomatic cohort first.
"It's not like we're looking for cancer that doesn't exist, that might develop," she said. "You're looking for a cancer that is there that you can't see."
The team is set to conduct a prospective study with a few hundred samples to further validate the test within both USC and the spinout CpG Diagnostics, which Salhia founded to commercialize OvaPrint. The company aims to start CLIA validation after the prospective study and offer the assay as a laboratory-developed test before eventually going through the US Food and Drug Administration's approval process, she said.
While Salhia couldn't speak to the exact cost of the test, once it's available, she said she believed it will be in line with what's currently on the market. Similar tests currently range from about $950 to around $5,000.
Liquid biopsy tests utilizing DNA methylation are the focus of multiple companies, including Grail, Delfi Diagnostics, and Guardant Health, and they cover a broad range of diseases such as lung cancer and colorectal cancer, but the emphasis on ovarian cancer alone — particularly HGSOC — differentiates the USC test from others, Salhia said.
Grail has offered its multi-cancer early detection Galleri test clinically since 2021, which in a recent study in the UK had a positive predictive value of about 76 percent and a negative predictive value of nearly 98 percent. The test can detect a signal shared by more than 50 types of cancer in a blood sample as well as predict the tissue or organ associated with the cancer signal, or the cancer signal origin.
In a prospective study published in Annals of Oncology in 2021, overall sensitivity of the Galleri test for ovarian cancer was 83 percent, with sensitivity for stage I ovarian cancer at 50 percent. Accuracy for the cancer signal origin was 70 percent in patients with ovarian cancer, a spokesperson from Grail noted.
There is also competition within the ovarian cancer testing space: Northern Ireland startup GenoMe Diagnostics is currently validating its OvaMe test that uses DNA methylation markers and digital droplet PCR technology to detect ovarian cancer, although the company is in its early stages. UK-based firm Angle, meantime, is developing a gene expression assay to determine whether a pelvic mass is benign or cancerous, while Mercy BioAnalytics is developing an extracellular vesicle-based test, the Mercy Halo Ovarian Cancer test, to distinguish between HGSOC and benign masses. In preliminary data presented at the American Society of Clinical Oncology annual meeting in 2022, the Halo test distinguished HGSOC from benign cancer with about 90 percent sensitivity and 98 percent specificity.
The USC researchers are also considering broadening their scope by utilizing their technology for minimal residual disease and other types of early cancer detection, but ovarian cancer is the main focus for both the research team and CpG Diagnostics, Salhia noted.