Rhodium HTS
Rhodium version 380E6-4
Standard protocol provided in rhodium suite, entitled "HTS"
Docking calculations were prepared as detailed in the PosePredictionProtocol.txt file.
The final scoring was based on a proprietary metric called Rhodium KRh, producing a predicted IC50 value reported in the LigandScore.csv file. The KRh metric is based on a linear correlation between Rhodium's docking score and IC50 values for all ligands in the BindingMOAD, for which an crystal structure and an IC50 value in any assay was available.
Importantly, the KRh metric adjusts the standard Rhodium docking score for molecular weight, in the spirit of a "ligand efficiency" metric, and furthermore accounts for unsatisfied hydrogen bond donors and acceptors in penalizing hydrogen bond donor and acceptor atoms with no partner including the ligand pose and protein residues within 4 Angstroms of the ligand pose.
Rhodium HTS Mode
OpenBabel 2.3.90, Rhodium380E6-4 (propietary), PyMol 1.3
Openbabel standard option of --gen3D to create 3d structures from provided sdf files
Openbabel obminimize -n 20000 -ff MMFF94 for each ligand
Multiple rigid conformers from obabel --confab --rcutoff 0.7 --ecutoff 4
Selected protein structure models were downloaded from the RCSB website.
Next, hydrogen atoms were added with PyMol 1.3.
When multiple side-chain conformers were present, the first or "A" conformer was selected. Others were deleted from the .pdb file.
The prepared proteins were aligned to the provided apo template using the alignment tools available in PyMol.
Multiple rigid conformers (determined from the confab conformer generator) were docked in each of the downloaded protein models.
Default parameters were used.
Poses were predicted based on the minimized scoring function.
This includes an initial grid search for poses wherein trial poses are placed on a 3D grid of pose centers with resolution 1.7 Angstrom, surrounding the surface of the protein model.
Seventy-two trial poses (covering the SO3 rotation group) of each conformer are placed on each point, and scored
with a high-speed low-resolution, long-range potential similar to that of I. Vakser, et al.
Approximately the best 1% of those poses are refined with a docking score, based roughly on the XSCORE method, using the Multi Directional
Search of V. Torczon. This minimizer accounts for (or takes advantage of) chaotic behavior commonly found in pose optimization.