Skip to main content

Table 5 Performance of various hybrid models developed using combination of descriptors b .

From: DrugMint: a webserver for predicting and designing of drug-like molecules

Model

Descriptors

Threshold

Sensitivity

Specificity

Accuracy

MCC

AUC

Hybrid-1

50

0

86.86

86.28

86.45

0.7

0.92

Hybrid-2

50

0

74.09

65

67.69

0.36

0.73

Hybrid-3

100

−0.1

92.43

87.99

89.3

0.77

0.95

Hybrid-4

296

0

90.57

89.86

90.07

0.78

0.96

Hybrid-5

22

0

87.75

84.68

85.59

0.69

0.9

  1. bHybrid-1: top 5 descriptors from each fingerprints based on their positive correlation against activity, Hybrid-2: top 5 descriptors from each fingerprints based on their negative correlation against activity, Hybrid-3: sum of descriptors from Hybrid-1 and Hybrid-2, Hybrid-4: sum of all 10 types of fingerprints after applying CfsSubsetEval algorithm, Hybrid-5: Running the CfsSubsetEval algorithm on the descriptors set of Hybrid-4 (296).