Classification ROC Curve. ROC curves of binomial logistic classifiers on the riboswitches of the test set (23 sequences) and their filtered random shuffles (151 sequences). Classifier features shown in legend. SPSSⒸwas used to estimate logistic regression coefficients derived from the training set consisted of the 29 riboswitches and 216 sequences as negative control. Resolution (0.01) was used to calculate true and false positive rates. Corresponding threshold for a given false positive rate was used to calculate the true positive rate. Values strictly higher than threshold were used to calculate true positive rates. In the legend, the area under the ROC curve corresponding to each classifier is shown.