Skip to main content

Table 1 Test set performances of random forest models trained on data generated by Nonpher, SAscore and DR approaches

From: Nonpher: computational method for design of hard-to-synthesize structures

Model

Acc (%)

SN (%)

SP (%)

AUC

Nonpher

89.6

93.8

77.0

0.94

SAscore

82.5

94.7

46.0

0.89

DR

46.0

30.8

91.5

0.60

  1. An accuracy (Acc), sensitivity (SN), specificity (SP) and an area under a ROC curve (AUC) are calculated as average values from five different random samples of the \( S_{train} \) data set