Skip to main content

Table 3 Confusion matrix for anti-HIV activity of purine nucleoside analogs in human PBM cells

From: Diverse models for anti-HIV activity of purine nucleoside analogs

Model

Description

Ranges

Number of compounds predicted

Sensitivity (%)

Specificity (%)

Non-error rate (%)

Overall accuracy of prediction (%)

MCC

Active

Inactive

Decision tree

Training set

Active

08

00

100

100

100

>99.9

1.00

Inactive

00

28

Tenfold cross-validated set

Active

06

02

75

93

84

89

0.68

Inactive

02

26

Random forest

 

Active

05

03

62.5

89

75.7

83

0.52

Inactive

03

25

Support vector machine

Training set

Active

04

02

66.7

100

83.3

93

0.78

Inactive

00

21

Test set

Active

01

01

50

86

68

78

0.36

Inactive

01

06

  1. The recognition rate of decision tree-, random forest-, and support vector machine-based models is also shown