Data-driven deep learning docking
Schrodinger/In-house Deep Learning
Assumed pH 4.5
Tautomers considered
Gasteiger charges
Proteins and Ligands were prepared by Maestro's prepwizard with default parameters.
Ligand conformational libraries were generated using ConfGen in Schrodinger.
num_conformers=200
Ligands in the Protein Data Bank were collected based on the FP2 score to targets. The ligand conformers of each target were aligned to templates. The in-house deep learning algorithms were utilized to rank those poses.
No
Yes
Deep-Learning-Package-D
AG/DG/TDL-BP/Schrodinger
Use AG, DG and/or TDL-BP with default parameters
Use PDBBind as the training data for these models
The features generated by Algebraic Graph, Differential geometry, and Algebraic topology scores are utilized in in-house design neural network models to predict the binding free energy
No
Yes