Fig. 3

Development of the cancer-specific detection model (model 3). a Heatmap for serum miRNA expression of patients with benign tumors or cancer. N = cancer, 333; benign tumors, 29. b PCA mapping for serum miRNA expression of patients with benign tumors or cancer. c Work flow of patients for development of prediction model 3. Serum samples were obtained from 543 subjects, including 320 patients with ovarian carcinoma, 66 with borderline tumors, 29 with benign tumors and 115 non-cancer controls (from non-cancer control B). The sample set was divided into two groups, the discovery set and validation set. d ROC curves for detecting cancer patients using miRNAs selected for prediction model 3. e Diagnostic performance of the selected nine miRNAs in the discovery set and validation set. f Cancer specificity of the diagnostic index using prediction model 3 in the validation set (ovarian cancer, 160; non-cancer [from non-cancer control B], 100; non-epithelial ovarian cancer, 13; borderline ovarian tumor, 33; and benign ovarian tumor, 14). Each diagnostic accuracy (%) is indicated. The p values were calculated by χ² test