group (methyl) at this position in active molecules. Hydrogen bond donor contour
was observed near the nitrogen of the amide and other near the cyclopropyl
group. A large hydrogen bond acceptor contour was observed near the oxygen of
the amide.
Wadood et al. generated structure-based pharmacophore model using the X-ray
crystal structure with PDBID 3O8A with 5-(2-methylbenzimidazol-1-yl)N-cyclopropylthiophene-2-carboxamide as the co-crystallized ligand [93]. This
pharmacophore model was used to identify molecules from ChemBridge database.
Eighty-seven molecules were identified using this model system, and the hits were
further screened using molecular docking and binding energy calculations using
GOLD, and generalized Born interaction energies, and binding affinity using MOE
docking software. Using these filters, twenty-five molecules with variable chemical
classes were identified.
Tseng et al. worked on 3D-QSAR pharmacophore generation and docking-based
pharmacophore development from a group of sixty-seven inhibitors of PfDHODH
belonging to different chemical classes [94]. The training set consisted of
thirty-eight compounds and the test set includes twenty-five molecules. The pharmacophoric features used were hydrogen bond donor (HD), hydrogen bond
acceptor (HA), hydrophobic group (H), and hydrophobic aromatic (HR) and were
used during HypoGen during hypothesis generation. Hypo1 pharmacophoric model
was considered to be the top model due to its high correlation coefficient (0.935),
lowest RMS deviation (2.15) and successful prediction efficiency of training
(89.4%) and test sets (72.4%). In docking-based pharmacophore generation,
sixty-seven molecules (both training and test set) were docked using X-ray crystal
structure of PfDHODH with PDB ID 1TV5. The pharmacophore models generated
were based on top scoring pose and genetic algorithm-based generation of 255
conformers. The scoring function was based on the interactions shown by the
molecules with His185, Arg265, and Tyr528 [94, 95]. The docking-based pharmacophore model, DBP-All255 (docking-based pharmacophore (DBP) of all 255
conformers) was found to show comparable results as that of Hypo1. In Hypo1, the
hydrophobic group feature was observed on the left side of the HA feature, and in
DBP-All255, the hydrophobic feature appears on the right side of the HA feature.
Both the models were able to predict the potential bioactive conformation of the
inhibitors based on the structure activity relationship and binding mode of the
inhibitors.
Hou et al. in 2016 performed QSAR on PfDHODH inhibitors using multilinear
regression (MLR) and support vector machine (SVM) [95]. A dataset of 255
molecules from ChEMBL database and literature, with PfDHODH activity, was
used. Most of the structures from the dataset contained triazolopyrimidine and
benzimidazole as basic moiety. 161 molecules as training set, 94 molecules as test
set, and 14 molecular descriptors (based on Pearson correlation) were selected. Four
final computational models were generated showing good prediction quality with q
2
(leave-one-out) > 0.66, correlation coefficient (r) > 0.85 on both training and test
sets. The mean square error (MSE) for training set is <0.32 and for test set is <0.37.
It was observed through this study that the antimalarial activity of the inhibitors is
Structure-Based Design of PfDHODH Inhibitors …
213
was observed near the nitrogen of the amide and other near the cyclopropyl
group. A large hydrogen bond acceptor contour was observed near the oxygen of
the amide.
Wadood et al. generated structure-based pharmacophore model using the X-ray
crystal structure with PDBID 3O8A with 5-(2-methylbenzimidazol-1-yl)N-cyclopropylthiophene-2-carboxamide as the co-crystallized ligand [93]. This
pharmacophore model was used to identify molecules from ChemBridge database.
Eighty-seven molecules were identified using this model system, and the hits were
further screened using molecular docking and binding energy calculations using
GOLD, and generalized Born interaction energies, and binding affinity using MOE
docking software. Using these filters, twenty-five molecules with variable chemical
classes were identified.
Tseng et al. worked on 3D-QSAR pharmacophore generation and docking-based
pharmacophore development from a group of sixty-seven inhibitors of PfDHODH
belonging to different chemical classes [94]. The training set consisted of
thirty-eight compounds and the test set includes twenty-five molecules. The pharmacophoric features used were hydrogen bond donor (HD), hydrogen bond
acceptor (HA), hydrophobic group (H), and hydrophobic aromatic (HR) and were
used during HypoGen during hypothesis generation. Hypo1 pharmacophoric model
was considered to be the top model due to its high correlation coefficient (0.935),
lowest RMS deviation (2.15) and successful prediction efficiency of training
(89.4%) and test sets (72.4%). In docking-based pharmacophore generation,
sixty-seven molecules (both training and test set) were docked using X-ray crystal
structure of PfDHODH with PDB ID 1TV5. The pharmacophore models generated
were based on top scoring pose and genetic algorithm-based generation of 255
conformers. The scoring function was based on the interactions shown by the
molecules with His185, Arg265, and Tyr528 [94, 95]. The docking-based pharmacophore model, DBP-All255 (docking-based pharmacophore (DBP) of all 255
conformers) was found to show comparable results as that of Hypo1. In Hypo1, the
hydrophobic group feature was observed on the left side of the HA feature, and in
DBP-All255, the hydrophobic feature appears on the right side of the HA feature.
Both the models were able to predict the potential bioactive conformation of the
inhibitors based on the structure activity relationship and binding mode of the
inhibitors.
Hou et al. in 2016 performed QSAR on PfDHODH inhibitors using multilinear
regression (MLR) and support vector machine (SVM) [95]. A dataset of 255
molecules from ChEMBL database and literature, with PfDHODH activity, was
used. Most of the structures from the dataset contained triazolopyrimidine and
benzimidazole as basic moiety. 161 molecules as training set, 94 molecules as test
set, and 14 molecular descriptors (based on Pearson correlation) were selected. Four
final computational models were generated showing good prediction quality with q
2
(leave-one-out) > 0.66, correlation coefficient (r) > 0.85 on both training and test
sets. The mean square error (MSE) for training set is <0.32 and for test set is <0.37.
It was observed through this study that the antimalarial activity of the inhibitors is
Structure-Based Design of PfDHODH Inhibitors …
213