Similarity/Alignment/Minimization
Rdkit/Torch/Macromodel
Assumed pH 7
Tautomers considered
OPLS3e charges
For each PDB entry identified as containing a reference ligand, any multiple chains
were aligned and binding geometries of multiple chains compared. In all cases the geometry differences were
minimal, so chain A was used. Maestro's prepwizard was then used to add hydrogens and assign bond orders using
using default parameters (pH 7). A restrained minimization using OPLS3e was also performed (converged at heavy
atom RMSD below 0.3 Angstrom). All the waters within 5 Angstroms of the ligand were retained. The ligands were
then split from the prepared proteins (and waters) and saved for use as an alignment references in Torch.
RDKit Morgen Fingerprint (radius=2, 1024 bits) Tanimoto similarities
Torch Conformational Hunt Setting - Very accurate and slow #generation of conformers for alignment
Torch Alignment Method - Maximum-common-substructure
Torch MCS Alignment Matching Rule - Very permissive
Torch MCS Alignment Excluded Volume Hardness - Medium
Macromodel force field OPLS3e
Macromodel unconstrained minimization of aligned ligand and all residues (and waters) with 3 Angstroms of the ligand
Macromodel constrained minimization of residues within 6 Angstrom of unconstrained atoms
For each of the 20 query ligands the most similar BACE-bound ligand in the PDB was identified as
an alignment reference using the RDkit Morgan fingerprints. Each query was aligned to its reference ligand using a maximum
common substructure alignment in Torch. The aligned query was then merged back into the reference protein structure (with waters)
and energy minimized.
Yes