Skip to main content
. 2022 Jan 10;126(3):472–481. doi: 10.1038/s41416-021-01593-6

Fig. 2. Optimisation of multi-miRNA biomarker panels.

Fig. 2

a Boxplots of AUC of multi-miRNA biomarker panels (with 2–12 miRNAs), in both the training and test sets, calculated from 200 iterations of the two-fold cross-validation process. The boxplot presents the 25th, 50th, and 75th percentiles in panel AUC. b Median AUC for the training and test sets from 200 iterations of the two-fold cross-validation process for multi-miRNA panels with 2–12 miRNAs. ***p < 0.001 (Student’s t-test). c ROC curves for breast cancer prediction performance of the optimal eight-miRNA biomarker panel in the Discovery and Validation 1 cohorts. The point with the maximum classification accuracy is shown as the red box. The sensitivity and specificity values at the maximum accuracy point are also shown. The 95% CI for these values is shown in the brackets.