Premium
Comparative QSAR Analyses of Competitive CYP2C9 Inhibitors using Three‐Dimensional Molecular Descriptors
Author(s) -
Lather Viney,
Fernandes Miguel X.
Publication year - 2011
Publication title -
chemical biology and drug design
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.59
H-Index - 77
eISSN - 1747-0285
pISSN - 1747-0277
DOI - 10.1111/j.1747-0285.2011.01106.x
Subject(s) - quantitative structure–activity relationship , molecular descriptor , conformational isomerism , loo , partial least squares regression , test set , chemistry , applicability domain , biological system , artificial intelligence , mathematics , molecule , computational chemistry , stereochemistry , computer science , machine learning , biology , organic chemistry
One of the biggest challenges in QSAR studies using three‐dimensional descriptors is to generate the bioactive conformation of the molecules. Comparative QSAR analyses have been performed on a dataset of 34 structurally diverse and competitive CYP2C9 inhibitors by generating their lowest energy conformers as well as additional multiple conformers for the calculation of molecular descriptors. Three‐dimensional descriptors accounting for the spatial characteristics of the molecules calculated using E‐Dragon were used as the independent variables. The robustness and the predictive performance of the developed models were verified using both the internal [leave‐one‐out (LOO)] and external statistical validation (test set of 12 inhibitors). The best models (MLR using GETAWAY descriptors and partial least squares using 3D‐MoRSE) were obtained by using the multiple conformers for the calculation of descriptors and were selected based upon the higher external prediction ( values of 0.65 and 0.63, respectively) and lower root mean square error of prediction (0.48 and 0.48, respectively). The predictive ability of the best model, i.e., MLR using GETAWAY descriptors was additionally verified on an external test set of quinoline‐4‐carboxamide analogs and resulted in an value of 0.6. These simple and alignment‐independent QSAR models offer the possibility to predict CYP2C9 inhibitory activity of chemically diverse ligands in the absence of X‐ray crystallographic information of target protein structure and can provide useful insights about the ADMET properties of candidate molecules in the early phases of drug discovery.