Skip to main content
. 2019 Sep 30;20(19):4855. doi: 10.3390/ijms20194855

Figure 2.

Figure 2

A contribution of the performance of prediction models with different wash conditions in the preparation of chemical structures of molecular operating environment (MOE) software. In the preparation for 3D chemical structures using MOE software, combinations of three kinds of protonation (none, dominate, and neutralize) and coordinates (2D, 3D, and CORINA) were utilized. The image produced by DeepSnap had the following angles and parameters: (176°, 176°, 176°), MPS:100, ZF:100, AT:23%, BR:14.5 mÅ, BMD:0.4 Å, BT:0.8 Å using nonoverlapped samples (Tra:Val:Test = 16:16:1) of the build DL-based prediction model by GoogLeNet from LR:0.001 to 0.0001 (a). The averages of AUCs of each LR were calculated (b).