Skip to main content

Table 3 Enrichment factor computed on the test set 1 (70 active molecules out of 701 compounds)

From: Convolutional architectures for virtual screening

Architecture EF 1% EF 2% EF 5% EF 10%
Tuned-MLP-Out 7(100%) 14(100%) 34(97.14%) 65(94.20%)
Voting 7(100%) 14(100%) 34(97.14%) 61(88.40%)
CNN 1D (M) 7(100%) 13(92.86%) 33(94.29%) 62(85.89%)
CNN 2D (R-M-F) 7(100%) 12(85.71%) 32(91.43%) 61(88.40%)
RF(F) 7(100%) 14(100%) 35(100%) 63(91.30%)
SVM(F) 7(100%) 14(100%) 35(100%) 61(88.40%)
  1. Fingerprint types: (R)DKIT,(M) organ, (F)eatMorgan,