Skip to main content

Table 4 Summary of performance metrics for each of the trained models

From: Mapping the stabilome: a novel computational method for classifying metabolic protein stability

Model AUC F-score MCC Sensitivity Specificity
   μ   σ   μ   σ   μ   σ   μ   σ   μ   σ 
BN+SVM(1) 0.85 0.0026 0.8 0.002 0.58 0.01 0.9 0.031 0.67 0.045
BN(1) 0.84 0.0012 0.8 0.0002 0.58 0.0007 0.9 0.005 0.66 0.0076
SVM(1) 0.74 0.0043 0.73 0.004 0.42 0.014 0.83 0.027 0.58 0.044
BN+SVM(2) 0.75 0.0052 0.77 0.009 0.39 0.035 0.88 0.032 0.46 0.072
BN(2) 0.66 0.0046 0.73 0.001 0.29 0.07 0.82 0.08 0.45 0.18
SVM(2) 0.73 0.0041 0.76 0.004 0.35 0.03 0.85 0.039 0.46 0.08
  1. Various performance metrics for each of the models trained on the full data set, as well as the trimmed data set. For example, BN+SVM(1) refers to the BN+SVM model trained on the full data set, while BN+SVM(2) refers to the same model trained on the trimmed data set. For each model we present the area under the curve (AUC) for a receiver operating characteristic analysis, and find the maximum F-score. The threshold from the maximum F-score was also used to calculate Matthews correlation coefficient (MCC) as well as sensitivity and specificity. For each metric the mean (μ) and standard deviation (σ) is shown for five cross-validation runs.