Molecular similarity/ ligand alignment
RDkit (2018.03.4) / MOE (2018.0101) / SWISS-MODEL
Default MOE parameters
None
Default MOE parameters
A total number of 269 BACE-1 complex structures were first downloaded from the PDBbind website. After removing the peptide-like ligands, the size of this dataset was reduced to 261, which we call the PBNP (PDBbind-BACE-Nonpeptide) dataset. The molecular similarities between the target ligands (D3R ligands) and the crystal ligands in the PBNP dataset were then calculated using the circular fingerprints (ECFP4) generated via RDKit. It was observed that one of the complex structures in the PBNP dataset, 3DV1, contains a macrocycle ligand which has a moderate to high similarity with most of the target ligands. Therefore, the A chain information of the receptor in this crystal structure was uploaded as the template to the SWISS-MODEL website to build the homology model of the target protein, based on the sequence information provided by the organizers. To make sure that the initial structures of the target ligands have good alignment with the crystal ligand in 3DV1, the MCS (maximum common substructure) algorithm in RDKit was used to generate 3D conformations of the target ligands from the SMILES string provided by the organizers, using the common substructures between the target ligands and 3DV1 ligand as templates. The PDB files of the homology model of the target protein and the 3D conformations of the target ligands were then joined together, in addition to the crystallographic waters preserved from 3DV1. Those file were then superimposed onto 5YGX-BACE_1-1.pdb provided by the organizers. After superposition, they were uploaded to MOE as starting structures for minimization and docking.
Yes
Yes