Skip to main content

Table 3 The AUC values of RF and SVM models constructed using different feature sets at the first stage

From: SIMLIN: a bioinformatics tool for prediction of S-sulphenylation in the human proteome based on multi-stage ensemble-learning models

Feature setsAUC
 RF
(class weight balanced)
SVM
(kernel function)
AAC0.680.63 (Polynomial kernel)
AAindex0.690.69 (Radial basis function kernel with grid search hyperparameter tuning)
ACC0.710.64 (Radial basis function kernel)
BINARY0.590.71 (Polynomial kernel)
BLOSUM620.680.74 (Radial basis function kernel)
CKSAAP0.660.63 (Polynomial kernel)
DISOPRED0.540.55 (Linear kernel)
PSIPRED0.620.60 (Polynomial kernel)
PSSM0.730.71 (Polynomial kernel)
Selected features
(mRMR+forward consequential elimination)
0.750.72 (Linear kernel)
  1. The bold font shows the highest performance of each feature among the RF and SVM