Skip to main content

Table 2 SCOP inter-version blind prediction using random forest

From: Automatic structure classification of small proteins using random forest

Shared SCOP Level

Evaluation Strategies (Accuracy)

 

Strategy 1 (85%)

Strategy 2 (93%)

 

Pre

Rec

MCC

Pre

Rec

MCC

Class

0.89

0.84

0.73

0.94

0.99

0.81

Fold

0.86

0.45

0.61

0.93

0.41

0.61

Super-family

0.80

0.55

0.69

0.75

0.55

0.63

Family

0.82

0.87

0.83

0.84

0.85

0.83

None

0.81

0.91

0.76

n/a

n/a

n/a

  1. The prediction accuracies and the class-wise performance of the random forest in strategies 1 and 2. Percent accuracy is the percentage of correctly classified domain pairs from all of the shared SCOP levels. Pre = Precision, Rec = Recall and MCC = Matthew's correlation coefficient.