Abstract

Ovarian cancer (OC) is the 5th leading cause of cancer-related death in women, partly due to difficulty in early diagnosis. Extracellular vesicles (EVs) show promise for use in early diagnostics of OC. EVs from cervical mucus (CM) of ovarian cancer patients were used for discovery of biomarkers for diagnostics. Machine learning was used to mine EV miRNA data to develop the panel.

EVs were extracted from the CM of 42 patients (26 tumor, 16 benign) for small RNA-sequencing. Both supervised and unsupervised approaches were applied and vetted against patient symptomology data. An independent ovarian cancer dataset was used for validation.

A biomarker panel of 9 microRNAs (voom: 96.55% and RF: 88% accuracy) was identified. Examination of the miRNA targets reveal that the panel is a good predictor as miRNA targets were functionally associated with pathways specific in OC progression.

Our method has identified EV miRNA biomarkers that could be crucial for early, non-invasive detection of OC. Data science has been used to develop a feedback system integrating experimental data and previously available data to identify a biomarker panel for OC diagnostics.