Expert opinion
VMD
none
Following the pose prediction work, the modelled ligands were visualised using the VMD
software. The literature was consulted (BMCL 19 (2009) 2595–2598; BMCL 21 (2011) 191–194;
BMC 21 (2011) 1134–1140 ) to identify patterns in known SARs. JM assigned binding energy estimate
for each ligand after qualitative evaluation of predicted poses, and consideration of literature SARs.
The uncertainty was 1 kcal/mol for most predictions, except for some cases where greater confidence was expected.
These cases were assigned a lower uncertainty of 0.6 kcal/mol.
All assignments were done on the 14th of October 2016, before any free energy calculations on the D3R sets were started.
Maestro/Marvisketch/rDock Visual
Maestro 11 (Beta Version)/ fconv/ Open Source Pymol 1.7 /MarvinSketch 15.3.30/ rDock
Assumed pH 7.4
Tautomers considered
Maestro's prepwizard was used to add hydrogens using default parameters.
The FXR structure provided by the organizers was used as initial template.
The structure was converted to mol2 files using fconv for the docking calculations, after initial cleaning with Maestro.
For the alchemical free energy calculations it was necessary to model a missing fragment comprised of residues A459-K464.
To this end the protein residues between N448 and Q476 were replaced by the same fragment as crystallized in the 3OKH structure. Subsequently, ACE capping groups were added to residues M247 of the main chain and D743 of the co-activator fragment.
Similarly an NME capping group was attached to D755 of the co activator fragment.
Ligand 3D structures were generated from 2D sdf files provided using MarvinTools scripts.
No water molecules were retained for docking calculations.
For the alchemical free energy calculations coordinates for water molecules accompanying the X-ray structure provided by
the organisers were superimposed with the coordinates of the poses. It was found that 1 water molecule
was susceptible of interfering with the simulations of the largest compounds (such as fxr_102). Consequently it was manually displaced to a nearby position.
RECEPTOR_FLEX 3.0 # Receptor flexibility weight
SITE_MAPPER RbtSphereSiteMapper # Cavity definition function
SMALL_SPHERE 1.5 # Small sphere radius for the cavity mapper
LARGE_SPHERE 4.0 # Large sphere radius for the cavity mapper
SCORING_FUNCTION RbtCavityGridSF #Cavity scoring function
WEIGHT 2 # Pharmacophoric restraint penalty weight
Docking calculations were performed for compounds fxr_91, fxr_101 and fxr_102. Docking was performed with rDock, generating the cavity using the two sphere method available in the program, centering a 15 A cavity within residues M294, I356, S336 and Y373 using 1.5 and 4.0 A for the radius of small and large spheres respectively. As pharmacophoric restraint an aromatic ring was forced within 4 A of the the center of the cavity.
Coincident binding modes were obtained for the 3 compounds. However, to minimize the differences between binding modes of the different compounds, poses for the 17 compounds in the set were derived from a common scaffold.
To this end, the BM of the largest compound FXR_102 was selected as a template and subsequently modified in Maestro to obtain poses that were minimized using the internal forcefield to avoid steric clashes.