As reported in Scientific Reports, Gaul and colleagues used high-performance mass spectrometry to interrogate the serum metabolome of patients with early-stage ovarian cancer and healthy controls. They were able to define a linear support vector machine model of 16 metabolites that identified early-stage ovarian cancer with high accuracy.
An initial support vector machine model using 255 identified metabolic features had moderate predictive accuracy (62%; 57% specificity, 67% sensitivity) in distinguishing samples from 46 patients with early-stage (I/II) serous epithelial ovarian cancer from those from 49 age-matched healthy controls. With use of a recursive feature method, a support vector machine–recursive feature method model was identified; it selected a minimum of 16 metabolic features that provided 100% accuracy (100% sensitivity, 100% specificity) in distinguishing between epithelial ovarian cancer and control samples. Many of the identified features in the support vector machine–recursive feature method model were lipids or fatty acids.
The investigators concluded: “The results provide evidence for the importance of lipid and fatty acid metabolism in ovarian cancer and serve as the foundation of a clinically significant diagnostic test.”