Skip to main content
. 2016 Aug 25;6:31479. doi: 10.1038/srep31479

Figure 3. Multi-class RF classification of diabetes subtypes.

Figure 3

Random Forest (RF) classification was performed in the R environment using the randomForest package. Data used was BMI, age, and gender-adjusted logFC for differentially abundant circulating miRNAs and the clinical designation (class label) of each individual [i.e., Healthy, Prediabetes, LADA, T2D, and T1D]. (A) Variable importance plot (Gini scores) determined during the initial multi-class RF classification including all 8 differentially abundant circulating miRNAs. This order of variable importance was used to recursively repeat the RF classification including the top 2, 3, and so forth combinations of miRNAs as predictor variables, and identify the binary classifier with the lower out-of-bag (OOB) estimate of error rate. Outline-colored box encloses the combination of miRNAs that generated the classifier with the lower OOB error rate (reported in the top left corner). (B) Multidimensional scaling (MDS) plot generated for the multi-class RF classifier that included the top 6 miRNAs and performed with the lower OOB estimate of error rate. (C) Receiver Operator Characteristic (ROC) Curves generated by plotting the classifier true positive rate (sensitivity) as a function of the false positive rate (1-specificity). The multi-class RF prediction probabilities were used for the generation of ROCR prediction objects. The 5-class classification problem was reformulated as five one-versus-all (OVA) binary comparisons. The area under the curve (AUC) is reported as performance measure. (D) Diagnostic Odds Ratios calculated for each OVA comparison using the formula DOR = (TP/FN)/(FP/TN), where TP is the number of true positives, FN, the number of false negatives, FP, the number of false positives, and TN, the number of true negatives. DOR of a test is a single indicator of diagnostic performance and represents the odds of positivity in individuals with the disease relative to the odds of positivity in individuals without the disease. Color and symbol coding: black and H: Healthy group; orange, P, and PreT2D: Prediabetes group; blue and L: LADA group; red and 2: T2D group; green and 1: T1D group.