Development of Diagnostic Biomarkers for Detecting Diabetic Retinopathy at Early Stages Using Quantitative Proteomics

Figure 9

A comparison of the discriminatory power of the 4-marker panel with the best single marker among the No DR, Mo NPDR, and Mi + Mo NPDR groups. We selected 4 proteins from t-test and stepwise MANOVA analyses and used them to construct a 4-marker panel. For comprehensive statistical analysis, LDA algorithms were employed where leave-one-out cross validation (LOOCV) was used to evaluate the discriminatory power among the No DR, Mo NPDR, and Mi + Mo NPDR groups. For comparisons between the best single marker and 4-marker panel, results are presented as confusion matrices. Sensitivity, specificity, error rates, and ROC curves are also represented with AUC values.

(a)

(b)