Identification of a blood based metabolomic classifer for the detection of ovarian cancer – MeDOC

Ovarian Cancer (OvCa) is the most lethal cancer among gynecologic malignancies. There are no sufficient screening options for early OvCa. Metabolomics is an approach to study small molecules, providing characteristic biochemical phenotypes in disease, and facilitates the development of novel diagnostic tools. Little is known about its potential use for OvCa early detection which was the aim of our study.

Blood samples from 407 individuals (Training: 100 controls, 33 OvCa, 80 Breast Cancer (BCa) Validation: 50 controls, 35 OvCa, 109 BCa) were collected and processed at the UFK Heidelberg. The endogenous metabolites in the plasma samples were analyzed with a targeted, quantitative, and quality controlled metabolomics approach by using the AbsoluteIDQ® p180 Kit (Biocrates). The analytical process was performed with the Biocrates MetIDQ™ software. Statistical analysis was done with R software. After data processing a random forest classifier was trained on the discovery dataset to separate OvCa from controls. Feature selection was done by elastic net regularization regression analysis on the training data resulting in the MeDOC classifier (5 single metabolites, 4 metabolite ratios).

The MeDOC classifier could differentiate OvCa patients from controls (88.5% sens.; 100% spec.); early stage OvCa (FIGO I+II) vs. controls (spec. 100%; sens. 90%) and late stage OvCa (FIGO III&IV) (88% sens.; 100% spec.). Further, the classifier could differentiate OvCa patients from BCa patients (sens. 97%).

In this study, we identified the MeDOC classifier with its metabolic markers as promising approach for the detection of OvCa. Multicentric, prospective studies are needed to evaluate its large-scale use.