Skip to main content

Table 1 Protein complex classification results obtained using classical SVM, Random Forest and XGBoost using input and privileged features with LOCO cross-validation over the affinity benchmark dataset

From: Learning protein binding affinity using privileged information

Features

Classical SVM

Random forest

XGBoost

ROC

PR

Sr

ROC

PR

Sr

ROC

PR

Sr

Input space

 2-mer

0.72

0.68

−0.40

0.68

0.63

− 0.38

0.72

0.66

−0.40

 Blosum (Protein)

0.70

0.63

−0.36

0.69

0.62

−0.39

0.69

0.63

−0.34

Privileged space

 NIRP

0.74

0.71

−0.45

0.74

0.67

−0.44

0.72

0.69

−0.42

 Moal descriptors

0.73

0.68

−0.43

0.70

0.68

−0.37

0.71

0.68

−0.34

 Dias descriptors

0.72

0.69

−0.42

0.69

0.69

−0.37

0.71

0.67

−0.34

 Blosum (Interface)

0.61

0.60

−0.19

0.56

0.54

−0.11

0.66

0.59

−0.25

  1. Bold faced values indicate best performance for each model. Blosum (Protein) refer to Blosum substitution scores averaged over the protein, while Blosum (Interface) are Blosum substitution scores averaged over the interface. Moal descriptors are taken from Moal et al. [8], and Dias descriptors are taken from Dias and Kolaczkowski [11]
  2. ROC Area under the ROC curve, PR Area under the precision-recall curve, Sr Spearman correlation coefficient