Detection of Endometrial Cancer With Proteomics & Machine Learning

Endometrial Cancer: Endometrial CancerResearchers have unveiled a promising approach to the early detection of endometrial cancer using proteomic analysis of cervico-vaginal fluid and blood plasma. The study leverages the power of proteomics and machine learning to identify unique protein signatures that can discriminate between endometrial cancer and non-cancerous conditions, offering a non-invasive and cost-effective alternative to current diagnostic methods. The study was published in eBioMedicine, part of The Lancet Discovery Science (2024; https://doi.org/10.1016/j.ebiom.2024.105064). Endometrial cancer, the most prevalent gynecological malignancy in high-income countries, has seen a steady rise in incidence, correlating with increasing obesity rates. Early diagnosis is crucial for favorable outcomes; however, current diagnostic procedures, including transvaginal ultrasound, hysteroscopy, and endometrial biopsy, are invasive and often distressing for patients. Moreover, only a small fraction of symptomatic women are found to have malignant conditions, underscoring the need for more efficient and less intrusive diagnostic tools. The study focused on symptomatic postmenopausal women, obtaining blood plasma and cervicovaginal fluid samples from 53 patients with endometrial cancer and 65 control subjects. Proteomic maps of these samples were created using Sequential Window Acquisition of All Theoretical Mass Spectra (SWATH-MS). Machine learning algorithms were then applied to identify the most discriminatory proteins, aiming to develop diagnostic models based on accuracy and simplicity. Key Findings The researchers made several discoveries regarding the effectiveness of protein signatures in distinguishing endometrial cancer from non-cancerous conditions. The study identified a five-protein biomarker panel from cervicovaginal fluid that demonstrated exceptional diagnostic accuracy. This panel included haptoglobin (HPT), galectin-3 binding protein (LG3BP), fibrinogen alpha chain (FGA), lymphocyte antigen 6 complex, locus D (LY6D), and immunoglobulin heavy constant mu (IGHM). This biomarker panel achieved an area under the curve (AUC) of 0.95 with a sensitivity of 91 percent and a specificity of 86 percent. This high level of accuracy suggests that cervico-vaginal fluid is a reliable medium for detecting endometrial cancer. A three-marker panel was derived from plasma samples, consisting of apolipoprotein D (APOD), proteasome subunit alpha type-7 (PSMA7), and HPT. The plasma biomarker panel achieved an AUC of 0.87 with a sensitivity of 75 percent and a specificity of 84 percent. While slightly less accurate than the cervicovaginal fluid panel, these results still highlight the potential of plasma biomarkers in cancer diagnostics. The diagnostic performance of the biomarker panels was particularly notable for early-stage (Stage I) endometrial cancer. The cervicovaginal fluid panel achieved an AUC of 0.92, indicating strong diagnostic capability for early detection. The plasma panel achieved an AUC of 0.88, reinforcing its utility in identifying early-stage cancer despite being less effective than the cervicovaginal fluid panel. To learn more about the study, Oncology Times chatted with first-author Kelechi Njoku, MBBS, MPH, MRCP (UK), PhD, a National Institute for Health and Care Research Academic Clinical Lecturer in Cancer Diagnostics and Machine Learning at the University of Manchester, U.K. Oncology Times: Can you elaborate on why cervicovaginal fluid was found to be a better medium for biomarker detection compared to blood plasma in this study? How can these biomarkers improve patient outcomes, especially in terms of treatment options and survival rates? Njoku: “Cervicovaginal fluid was a better medium for endometrial cancer biomarker detection compared to plasma due to the anatomical continuity of the uterine cavity with the lower genital tract which allows for cancer derived biomolecules in the uterine cavity to be shed down the female genital tract and collected by gentle lavage in the cervicovaginal fluid. Proximal fluids such as endometrial secretions derive from or are in direct contact with the tumor and thus are viable sources of cancer biomarkers. Blood derived biomarkers on the other hand, are limited by dilution, which explains the low yield of cancer-derived biomarkers in blood, especially in early-stage cancers. “In current clinical practice, symptomatic women with suspected endometrial cancer undergo sequential investigations that are invasive, painful, and anxiety provoking. For most, these are not needed as only 5-10 percent of symptomatic women have a sinister underlying pathology. Our cervicovaginal fluid assay is simple, safe, patient-friendly, and highly accurate and showed potential for detecting early-stage, advanced-stage, and biologically aggressive endometrial cancers. If translated into routine clinical care, it would improve the detection of endometrial cancer and could even be used by practice nurses and clinicians in community health care settings. Furthermore, it would enable women to collect their own vaginal fluid sample at a time and location that is convenient for them, thus extending its potential utility to high-risk asymptomatic women for whom no evidence-based screening test currently exists.” Oncology Times: What kind of confirmatory studies or clinical trials are planned to validate these biomarker signatures, and what key metrics will determine their success? Njoku: “Our next goal is to validate the 5-marker panel of proteins in a larger independent cohort of women with or at risk of endometrial cancer. We have recently received funding support from two leading UK charities—the Eve Appeal and the North West Cancer Research—to do this. We would be recruiting 500 symptomatic women for the confirmatory study, including those with a confirmed diagnosis of endometrial cancer, as well as those who do not have endometrial cancer to serve as controls. “We are looking at developing clinically tractable assays using established platforms such as ELISA, Lumipulse technology, or even lateral flow test for point-of-care testing. Our key metric for success would be achieving as high enough sensitivity and specificity as possible, ideally over 90 percent. Our goal is to develop a simple, patient-friendly ‘rule-out’ test that can enable quick reassurance for many symptomatic women who do not have cancer, and urgent investigations for those who test positive.” Oncology Times: How can this proteomic biomarker discovery be integrated with other diagnostic modalities, such as imaging and genetic testing, to enhance the overall accuracy and reliability of endometrial cancer diagnosis? Njoku: “This novel test could be used alongside transvaginal ultrasound scan to better streamline the diagnostic process for endometrial cancer. This would help reduce the number of symptomatic women who undergo invasive procedures, such as hysteroscopy and biopsy, that are unwarranted.” Dibash Kumar Das is a contributing writer.