Early Detection of Ovarian Cancer May be Possible with Combination of Artificial Intelligence and Liquid Biopsies

A liquid biopsy – a blood test – in combination with machine learning/artificial intelligence (AI) to detect and analyze distinctive tumor components, cancer-related genetic changes, and protein biomarkers released into the peripheral circulation, could help screen women for early signs of ovarian cancer, enabling a new accessible approach for non-invasive ovarian cancer screening and diagnostic evaluation.

This conclusion is based on a study by a team of researchers at the Johns Hopkins Kimmel Cancer Center in collaboration with several other institutions in the United States and Europe.

The results of the study were published in the September 30, 2024 issue of Cancer Discovery, a journal of the American Association for Cancer Research (AACR). In the article, the researchers describe how they used AI-powered analyses of DNA fragments and two protein biomarkers to identify women with ovarian cancer, a group of diseases that originates in the ovaries, fallopian tubes, or peritoneum and come in a variety of tumor types. [1]

Ovarian Cancer Biomarkers
Conventional tissue biopsy methods and serological biomarkers such as cancer antigen 125 (CA-125) and human epididymis protein 4 (HE4) have limited clinical applications. Although these two protein biomarkers were previously identified as ovarian cancer biomarkers, on their own, they cannot reliably detect ovarian cancer. However, combining these biomarkers with AI-driven detection of cancer-associated patterns of DNA fragments (cell-free DNA or cfDNA) in the circulation, improved screening accuracy and helped distinguish cancerous tumors from benign growths.

Results from a previous retrospective study, presented at the American Association for Cancer Research (AACR) Annual Meeting, held April 5-10, 2024, suggested that this approach could differentiate patients with ovarian cancer from healthy controls or patients with benign ovarian masses.

“The combination of artificial intelligence, whole-genome cell-free DNA (cfDNA) fragmentomes and, a pair of protein biomarkers (CA-125 and HE4) in a simple blood test improved detection of ovarian cancer even in patients with early-stage disease,” noted Victor E. Velculescu, M.D., Ph.D., senior author of the study, professor of oncology, and co-director of the Cancer Genetics and Epigenetics Program at the Johns Hopkins Kimmel Cancer Center.

“This AI-enabled approach has the potential to be an affordable, accessible method for widespread screening for ovarian cancer,” Velculescu added.

A common cancer
According to recent data from the Centers for Disease Control and Prevention (CDC), ovarian cancer is the fifth most common cause of cancer deaths among women in the United States.

The majority of patients with ovarian cancer are diagnosed in a late stage with advanced disease. As a result, this cancer has the highest mortality amongst of all gynecological malignancies in the US and Europe with a 5-year survival rate of approximately 51%. In contrast, for early-stage disease, the 5-year survival rate is above 90%.

One of the primary reasons of late stage discovery of ovarian cancer is that in most cases patients with earlier stages the disease are often are asymptomatic or present with unspecific symptoms. In addition, there is no test for ovarian cancer in women without symptoms. Hence, there is a significant unmet medical need to for new diagnostic tools for early diagnosis.

Saving lives
“Early detection of ovarian cancer may save lives but most women are diagnosed late in the course of the disease when survival rates are much lower,” explains co-first author Jamie Medina, Ph.D., postdoctoral fellow at the Johns Hopkins Kimmel Cancer Center.

“The lack of specific symptoms early in the course of the disease or effective biomarkers has hindered earlier detection efforts,” Medina added.

Liquid Biopsies and AI 
Liquid biopsy are minimal invasive diagnostic tests in which researchers analyze a patients’ blood for evidence of tumor-derived DNA shed from tumor cells for real-time detection of cancer. The overall advantages of liquid biopsies is that they are non-invasive, which allow for serial sampling and longitudinal monitoring of dynamic tumor changes over time.[2][3]

Although this approach has been explored as a way to non-invasively detect a variety of cancers, they have not always been useful in the detection and diagnosis of ovarian cancer. However, increasing evidence from new and ongoing studies seem to suggest that liquid biopsies may enhance the diagnostics and clinical management of ovarian cancer by improving early diagnosis, predicting prognosis, detecting recurrence, and monitoring response to treatment. In turn, capturing the unique tumor genetic landscape can also guide treatment decisions and the selection of appropriate targeted therapies. [2][3]

The investigators of this study previously demonstrated that the machine learning/artificial intelligence (AI-) powered DELFI (DNA Evaluation of Fragments for early Interception) test-method utilizes a new approach for liquid biopsies, measuring changes in genome-wide cell-free DNA fragmentation profiles in peripheral blood to reflect genomic and chromatin characteristics of lung cancer. The technology, called fragmentomics, improves detection of DNA fragments in the blood and effectively detects lung cancer.[4]

Based on the results from this unrelated prospective study, published June 3, 2024 in Cancer Discovery, the researchers were able to confirm that machine learning/artificial intelligence (AI) could identify people more likely to have lung cancer based on DNA fragment patterns in the blood. The study enrolled about 1,000 participants with and without cancer who met the criteria for traditional lung cancer screening with low-dose computed tomography (CT). The researchers believe that this novel approach may be  helping to identify patients most at risk and who would benefit from follow-up CT screening, this new blood test could potentially boost lung cancer screening and reduce death rates, according to computer modeling by the team.

The technology takes advantage of the fact that DNA, neatly packaged in healthy cells, becomes disorganized in cancer cells. When healthy cells die and break apart, they leave behind a predictable, orderly set of DNA fragments in the blood. However, when cancer cells die and break apart, the DNA fragments left behind are irregular and chaotic.[4][5]

Study approach
The latest study used blood samples from 94 women with ovarian cancer, 203 women with benign ovarian tumors, and 182 women without any known ovarian growths. The study population used to develop the approach comprised women treated at hospitals in the Netherlands and Denmark.

The researchers used the DELFI-Pro test, which combines AI-powered cell-free DNA analysis with tests for CA-125 and HE4, to analyze the samples for ovarian cancer screening.

The DELFI-Pro test was able to detect substantially more cases of ovarian cancer than tests for either protein alone, and it did so with almost no false positives. In fact, it detected 72%, 69%, 87%, and 100% of ovarian cancer cases stages I–IV, respectively, while at the same specificity, CA-125 alone detected 34%, 62%, 63%, and 100% of ovarian cancers for stages I–IV.

According to the authors of the study, their approach differentiated benign masses from ovarian cancers with high accuracy (AUC=0.88, 95% CI=0.83-0.92). These results were validated in an independent population.

These findings show that integrated cfDNA fragmentome and protein analyses detect ovarian cancers with high performance, enabling a new accessible approach for noninvasive ovarian cancer screening and diagnostic evaluation.

A second sample
To validate the results, the researchers used the test in a second sample of American women that included 40 patients with ovarian cancer, 50 patients with benign ovarian growths, and 22 without known ovarian lesions. Even in this smaller sample, the test achieved similar success rates, with 73% of all cancers detected and 81% of the high-grade serous ovarian carcinoma, the most aggressive form of the disease, with almost no false positives in women without cancer. The DELFI-Pro test was also able to effectively distinguish between benign growths and cancerous tumors — something ultrasound exams cannot.

“Ovarian cancers have a unique DNA fragmentation signature that is not present in benign lesions,” explained Akshaya Annapragada, co-first author and an M.D./Ph.D. student at the Johns Hopkins University School of Medicine.

Being able to distinguish benign from cancerous ovarian growth is important because the next step in cancer screening for women with ovarian growths detected via ultrasound is exploratory surgery. Using the liquid biopsy tests could spare women with benign growths having to undergo unnecessary surgery.

Velculescu and his colleagues intend to validate the test’s utility in larger samples from randomized clinical trials but he found the current results encouraging.

“This study provides further evidence demonstrating the benefit of genome-wide, cell-free DNA fragmentation and artificial intelligence to detect cancers with high accuracy. Our results show that that this combined approach has higher performance for screening than existing biomarkers,” Velculescu concluded.

Reference
[1] Medina JE, Annapragada AV, Lof P, Short S, Bartolomucci AL, Mathios D, Koul S, Niknafs N, Noe M, Foda ZH, Bruhm DC, Hruban C, Vulpescu NA, Jung E, Dua R, Canzoniero JV, Cristiano S, Adleff V, Symecko H, van den Broek D, Sokoll LJ, Baylin SB, Press MF, Slamon DJ, Konecny GE, Therkildsen C, Carvalho B, Meijer GA, Andersen CL, Domchek SM, Drapkin R, Scharpf RB, Phallen J, Lok CAR, Velculescu VE. Early detection of ovarian cancer using cell-free DNA fragmentomes and protein biomarkers. Cancer Discov. 2024 Sep 30. doi: 10.1158/2159-8290.CD-24-0393. Epub ahead of print. PMID: 39345137.
[2] Zhu JW, Charkhchi P, Akbari MR. Potential clinical utility of liquid biopsies in ovarian cancer. Mol Cancer. 2022 May 11;21(1):114. doi: 10.1186/s12943-022-01588-8. PMID: 35545786; PMCID: PMC9092780.
[3] Terp SK, Stoico MP, Dybkær K, Pedersen IS. Early diagnosis of ovarian cancer based on methylation profiles in peripheral blood cell-free DNA: a systematic review. Clin Epigenetics. 2023 Feb 14;15(1):24. doi: 10.1186/s13148-023-01440-w. PMID: 36788585; PMCID: PMC9926627.
[4] Mazzone PJ, Bach PB, Carey J, Schonewolf CA, Bognar K, Ahluwalia MS, Cruz-Correa M, Gierada D, Kotagiri S, Lloyd K, Maldonado F, Ortendahl JD, Sequist LV, Silvestri GA, Tanner N, Thompson JC, Vachani A, Wong KK, Zaidi AH, Catallini J, Gershman A, Lumbard K, Millberg LK, Nawrocki J, Portwood C, Rangnekar A, Sheridan CC, Trivedi N, Wu T, Zong Y, Cotton L, Ryan A, Cisar C, Leal A, Dracopoli NC, Scharpf RB, Velculescu VE, Pike LRG. Clinical validation of a cell-free DNA fragmentome assay for augmentation of lung cancer early detection. Cancer Discov. 2024 Jun 3. doi: 10.1158/2159-8290.CD-24-0519. Epub ahead of print. PMID: 38829053.
[5] Cristiano S, Leal A, Phallen J, Fiksel J, Adleff V, Bruhm DC, Jensen SØ, Medina JE, Hruban C, White JR, Palsgrove DN, Niknafs N, Anagnostou V, Forde P, Naidoo J, Marrone K, Brahmer J, Woodward BD, Husain H, van Rooijen KL, Ørntoft MW, Madsen AH, van de Velde CJH, Verheij M, Cats A, Punt CJA, Vink GR, van Grieken NCT, Koopman M, Fijneman RJA, Johansen JS, Nielsen HJ, Meijer GA, Andersen CL, Scharpf RB, Velculescu VE. Genome-wide cell-free DNA fragmentation in patients with cancer. Nature. 2019 Jun;570(7761):385-389. doi: 10.1038/s41586-019-1272-6. Epub 2019 May 29. PMID: 31142840; PMCID: PMC6774252.

Featured image: © 2024 Carolyn Hruban, Ph.D./ Johns Hopkins Medicine. Used with permission.

DOI: 10.14229/onco.2024.09.30.001