multitarget/Autodock_Vina/Pose&Rank
OpenBabel 2.4.0/VMD 1.9.2/pdb-tools (J. Rodrigues)/MGLTools/Autodock Tools 1.5.6/Autodock Vina 1.1.2/GROMACS 2018/Pose&Rank web server (Sali lab)
Assumed pH 5.1
Gasteiger charges
Protein: We collected all BACE-1 structures with resolution <=2.0 Å in the PDB and selected the subset of these structures that also occurred in the Platinum database for ligands. This subset contained the following 68 structures:
2qmf 2qp8 2vie 2vij 2viy 2vj7 2vj9 2xfj 3in3 3ine 3inf 3inh 3ivh 3k5g 3kn0 3l58 3ooz 3tpp 3udm 3udp 3udr 3udy 3ufl 3veu 4acu 4acx 4b70 4b77 4b78 4d88 4di2 4ewo 4exg 4fm7 4fm8 4frj 4fs4 4i0d 4i0f 4i0g 4i0j 4i11 4i1c 4j0p 4j0y 4j17 4j1e 4jp9 4jpc 4jpe 4ke1 4l7j 4lc7 4lxa 4n00 4pzw 4r5n 4rcd 4rcf 4rrn 4rro 4wy1 4x7i 4xkx 4xxs 4ybi 4zsp 4zsr
The first chain of each structure was used. Sequence alignment of the FASTA sequences with the one provided in the D3R initial package showed only minor differences (11 of the structures, maximum 6 residues) so no back-mutations were done. The structures were aligned to the reference structure using VMD (after renumbering the sequences so that the EMVDN subsequence starts at 62).
The box center and size were determined by finding the min and max coordinates of all bound ligands. Finally, all ligands were removed and the protein structures were converted to PDBQT files using MGLTools with default parameters.
Ligands:OpenBabel was used to protonate and geometry-optimize the ligand structures (obconformer 500 100). The secondary amines were assumed to be protonated (positively charged) except when neighboring an aromatic ring, in which case it was unprotonated (this is an uncertain assumption as pH is close to pKa).
The ligands were converted to PDBQT format by assigning charges and atom types using MGLTools and default parameters.
Molecular docking with rigid protein and flexible ligands
Number of targets: 68
Iterated Local Search global optimizer search method
Box size: 22 * 15 * 17 Å
Box center: X=22.87 Å, Y=59.54 Å, Z=88.59 Å (in the reference frame of 2qmf)
Exhaustiveness=200 #exhaustiveness of global search
Vina scoring funtion (empirical + knowledge-based function)
Best poses rescored using PoseScore (atomic distance-dependent statistical scoring)
Docking runs were executed for each of the 68 targets with the above specified parameters. All obtained poses were clustered using GROMACS tools (gmx cluster -method gromos -cutoff 0.15) and the clusters were sorted according to the lowest Vina score of any member of the cluster.
The five best clusters (represented by the lowest-score member) were submitted for rescoring with the PoseScore scoring function (H. Fan et. al, J.Chem.Inf.Model. 2011, 51:3078), using the Pose&Rank web server at A. Sali's lab:
https://modbase.compbio.ucsf.edu/poseandrank/
The rescoring did not affect the coordinates but only changed the internal ordering of the five poses.
No
Yes