Computational protocol: Searching for bioactive conformations of drug-like ligands with current force fields: how good are we?

Similar protocols

Protocol publication

[…] Hydrogen atoms were added to ligands based on the ligand templates from the curated library of ligands [, ] (Accessed: November 28, 2015). The most probable protonation state for each ligand was predicted by Epik [], at the pH used for crystallization (information on pH was read from the PDB file header), using water as solvent. Ligands containing multiple charged groups, with the total charge smaller than −3 or larger than +3 were excluded from any further analyses.Drug-likeness of ligands within this study was evaluated based on descriptors proposed by Lipinski [] and Veber [], allowing for some tolerances for upper and lower limits. Ligand efficiency or potency descriptors could not be used for defining drug-likeness, as there are no standardized data available. All ligands stem from crystal structures so their sufficient water solubility is assumed. Number of rotatable bonds (#RotB), molecular weight (MW), and molecular surface areas (3D: MSA, PSA) were calculated by our program and crosschecked with the values provided by FaF-Drugs2 [] (for #RotB) and QikProp []. Octanol–water partition coefficient (LogP o/w) was estimated and the number of heavy atoms (#HA) was calculated by QikProp. Polar surface area (2D-PSA) was computed by OpenBabel [, ], where the total PSA equals to the sum of atomic contributions. Descriptors, which were used for elimination and their threshold values, can be found in Additional file . [...] A two-step conformational search was carried out for the final set of ligands by MacroModel []. The first step was incorporated into the search protocol with the aim to remove possible positive structural bias associated with the fact that the crystal conformation was used as input. In the first step, choosing OPLS_2005 [] as the force field and GB/SA water model for the implicit solvation, a pool of conformers was generated for each ligand with mixed MCMM/Low-Mode conformational algorithm, by accepting all conformers with energies up to 50.0 kcal/mol from the global minimum. The maximum number of Monte Carlo steps was set to 5000. At maximum, 1000 conformers were saved for every individual ligand.Next, heavy-atom RMSDs were calculated between each conformer and the original crystal conformation (n.b. all RMSD calculations in this study are symmetry corrected, taking into account the structural automorphism; e.g. in case of a molecule containing a para-substituted benzene ring, the RMSD between two conformations is zero if these two conformations differ only by a 180° flip of the ring around the linker bond). The conformer with the highest RMSD from the crystal conformation, i.e. the most dissimilar one, was saved to a separate file and used as the input for the second step of conformational search.The second—production—step of conformational search was performed using four different force fields—OPLS3 [], OPLS_2005 [], MMFF94s [] and AMBER* [–], with the same operating parameter settings as in the first step, except that the energy window was set to 5.0 kcal/mol, as in this step the low-energy conformers were of interest.To evaluate the effect of different solvents, conformational searches were also done using octanol and chloroform. This solvent effect analysis follows the assumption that, e.g. a polar compound in a relatively hydrophobic protein binding site may adopt a collapsed bioactive conformer (self-folded, stabilized by intramolecular interactions), which would be probably more similar to conformers optimized using the octanol or chloroform as solvent. If optimized in water, such a polar compound would benefit from favorable interactions with the solvent environment and its conformers would extend to maximize the polar interactions, resulting in a conformer pool with members likely dissimilar from the bioactive conformer.Using only heavy atoms, RMSD values were calculated between all conformers found in the production step and the original crystal conformation. The conformer with the lowest RMSD from the crystal conformation, i.e. the most similar one, was used in the following analyses and statistical comparisons. […]

Pipeline specifications

Software tools FAF-Drugs, QikProp, Open Babel, OPLS3
Application Drug design