3D similarity pose-picking
Smina/Data-Warrior4.7.2
vinardo scoring function was used in the pose generation
First, the 154 test ligands were compared to the 20 published BACE ligands to find the most similar pair. Second, smina vinardo was excuted to dock the 154 test ligands.
Vinardo scoring function
add_box=2 #padding space of the ligand
Other parameters were set as default
Third, the most 3D similar pose (a python script was written to compute each pose's electron residue to determine the most 3D alike pose) was selected for affinity prediction.
No
Yes
ViChMLP
Smina/Data-Warrior4.7.2/PycharmCE4.5/sklearn
11 vina terms from predicted poses, 24 chemical properties, MLPRegressor with 10,8,8 layers with alpha = 7
First, the 154 test ligands were compared to the 20 published BACE ligands to find the most similar pair.
Second, smina vinardo was excuted to dock the 154 test ligands, and the most 3D similar pose (a python script was written to compute each pose's electron residue to determine the most 3D alike pose) was selected.
Thrid, MLPRegressor in sklearn was applied to predict the affinity based 11 vina terms (from simina) and 24 chemical properties (from Data-Warrior). This model was trained by 222 BACE affinity dataset in the PDBbind database.
No
Yes