Skip to main content

Table 5 Performance of the Tuned-MLP-Out network on three data sets with 1%, 2%, and 5% active/inactive proportion respectively

From: Convolutional architectures for virtual screening

Active/inactive proportion Bal.Accuracy Sensitivity Loss AUC F1-score MCC
1% 0.7475 0.5000 0.5116 0.9700 0.5333 0.5289
2% 0.9671 0.9375 0.5114 0.9415 0.9009 0.8226
5% 0.9382 0.8780 0.0565 0.9991 0.9230 0.9196