In a recent study published in eBioMedicine, researchers evaluated proteomic signatures in blood plasma and cervicovaginal fluid for endometrial cancer detection.
Study: Detection of endometrial cancer in cervico-vaginal fluid and blood plasma: leveraging proteomics and machine learning for biomarker discovery. Image Credit: mi_viri/Shutterstock.com
Background
Endometrial cancer stands as the most prevalent gynecological cancer in high-income countries, with increasing incidence rates linked to the global rise in obesity. Despite its high treatability when detected early, the prognosis for advanced stages remains grim, emphasizing the critical need for early detection.
Traditional diagnostic methods, involving transvaginal ultrasound, hysteroscopy, and biopsy, are effective yet invasive and stressful, driving the demand for simpler, non-invasive testing methods. Blood tests, while accessible, often fall short in early cancer detection due to low biomarker concentrations, particularly in small or early-stage tumors.
Conversely, cervico-vaginal fluid presents a promising alternative, leveraging the uterine cavity's connection to the lower genital tract to offer a less invasive biomarker source.
Past studies, including O'Flynn et al.'s work, have shown the feasibility of detecting endometrial cancer through cervico-vaginal fluid cytology, although challenges in reproducibility and the need for specialized expertise limit its practicality.
The advent of high-throughput technologies and artificial intelligence in proteomics now heralds a new era in cancer biomarker discovery, with platforms like SWATH-MS offering precise and reliable protein signature analysis.
This study aims to explore the diagnostic capabilities of cervico-vaginal fluid protein signatures, comparing their effectiveness against plasma-derived markers, and assessing their potential to identify not only early-stage but also advanced and aggressive forms of endometrial cancer.
About the study
In the present study, researchers evaluated the performance of proteomic signatures from cervicovaginal fluid and plasma for endometrial cancer detection.
They enrolled females with postmenopausal bleeding and those with endometrial cancer. Cases were females with histopathological evidence of endometrial cancer based on hysterectomy.
Controls were symptomatic females without endometrial cancer or atypical hyperplasia. Individuals with a history of gynecological malignancy or hysterectomy were excluded.
Cervicovaginal fluid and blood were collected, and mass spectrometry was performed. Digitized proteomic maps were derived using sequential window acquisition of all theoretical mass spectra.
The resultant spectral data were converted and searched against a human plasma library and a previously published library of 19,394 peptides and 2,425 proteins in the cervicovaginal fluid.
Random forest (RF) modeling was used for feature selection. The most discriminatory proteins were ranked based on the mean decrease in accuracy.
Nested logistic regression models were built by sequential addition of proteins based on their rank. The parsimonious model was identified. Model performance was evaluated by plotting the receiver operating characteristic curve and calculating the area under the curve (AUC).
Likelihood ratio tests and Akaike information criteria (AIC) were used to compare the performance of nested models.
Findings
Overall, 118 postmenopausal females with symptoms were recruited. Of these, 53 had confirmed endometrial cancer, and 65 had no evidence of cancer. Most participants (86%) were White, and those with endometrial cancer were likely to be older and have a higher body mass index (BMI) than controls.
In total, 597, 310, and 533 proteins were quantified in the cervicovaginal fluid supernatant, cell pellets, and plasma samples, respectively.
Overall, 941 unique proteins were identified across sample types. There was evidence of separation between cancers and controls based on cervicovaginal fluid supernatant proteins.
Classifiers were selected based on the mean decrease accuracy metric of the RF model. Principal component analyses (PCA) using the top discriminatory proteins revealed stronger discrimination between cancers and controls.
The model with the top five discriminatory proteins had the least AIC value and was selected as a parsimonious model.
This model predicted endometrial cancer with AUC, sensitivity, and specificity of 0.95, 91%, and 86%, respectively. Feature selection analysis indicated that 38 proteins were important for discrimination between cancers and controls.
Proteins in cervicovaginal fluid cell pellets were less promising as cancer biomarkers than supernatant-derived proteins.
There were fewer differentially expressed proteins in plasma between cases and controls compared to the cervicovaginal fluid, with little evidence of discrimination based on plasma proteins.
PCA indicated a modest separation between cancers and controls. A three-plasma biomarker panel predicted endometrial cancer with AUC, sensitivity, and specificity of 0.87, 75%, and 84%, respectively.
Feature selection analysis revealed six plasma proteins as important classifiers. Further, three- and four-marker panels of cervicovaginal fluid and plasma proteins predicted early-stage endometrial cancer with AUCs of 0.92 and 0.88, respectively.
Besides, five- and six-marker panels of cervicovaginal fluid and plasma proteins predicted advanced-stage endometrial cancer with AUCs of 0.96 and 0.93, respectively.
Conclusions
The increasing incidence of endometrial cancer in high-income countries, paralleled by rising obesity rates, underscores the urgent need for effective early detection methods.
The study highlights the potential of cervico-vaginal fluid and plasma as sources for minimally invasive detection of cancer-derived biomarkers, with a focus on proteins such as LG3BP and LY6D, which are linked to cancer processes and show promise in diagnosing endometrial cancer.
Immunoglobulins and other biomarkers have also been identified, pointing to the body's immunological response to malignancies. The research supports the feasibility of using high-throughput technologies to detect endometrial cancer through cervico-vaginal fluid proteins, which offer a more patient-friendly and accessible method compared to traditional approaches.
These findings, alongside the development of new diagnostic tools like the PapSEEK test, suggest that targeted proteomic analysis and biomarker panels could significantly improve early detection and treatment outcomes for endometrial cancer.
However, the study calls for further validation of these biomarkers and the development of clinically actionable assays, potentially transforming patient care through innovative diagnostic tests.
Study shows AI can predict prognosis in triple-negative breast cancer